If you found this file in an archive then use keyword "nutteingy2" in a
search engine to find an updated version or related pages.
Updated file November 2005
Articles posted to moderated Yahoo community group for forensic science
personel starting 16 July 2003,title of thread "Problems with DNA Profiles" under
my nom de guerre of Nona Revers or as it turns up there nonarevers.
Text in brown is my contribution - due to the nature of such dialogues the thread can get rather broken as it splits into different sub-themes at different times and involving different people.
Yahoo forensic science group A more
searchable archive of that group
Subject: Problems with DNA profiles
The following like the first "y" file can get a bit jumbled,
and probably sections missing,
for anyone wanting the original question/reply structure
then please consult the archives on
http://infoarchive.net/sgroup/forensic-science/
Two questions on DNA profiles
A crime-scene profile has a full set of 13 loci
so 26 data points. On 25 there is a perfect match
with a profile on a database. The 26th allele from
the crime is 18 but the database is 18.2 - is
this declared a match ?
Second question concerning another crime-scene
profile of a totally unknown/unwitnessed offender.
Frequency analysis of the alleles indicates he is
50 times more likely to have ethnicity A than B.
But one pair of alleles are extremely rare in A population
say 0.002,0.002 but in population B it is say 0.015,0.15 .
Do you declare no result or does one determination
override the other ?. As equally rare for both alleles
then it is difficult to argue mixed parentage
from A and B populations.
On the first part, I don't think anyone would declare it a match
without further examination. Even after further examination I still
think they wouldn't use a term like that.
For your second question, I'm not aware of anyone in the US that uses
the population databases to determine possible racial information.
On the first part a match should not be called, but the similarity between the
crime scene profile and that of the person on the database would probably
result in further tests, certainly at the locus in question. If these confirm
the discrepancy then that person should be eliminated, but I would expect
investigators to consider close relatives, especially twins of the same sex as
the person on the database.
Ancestral background is just when it all starts
getting interesting. I have been trying to establish
whether D2S1338 alleles 17 to 20 relates to
Anglo-Saxon (Germany ,Flemish,Saxony)
background and 20 to 24 a Celtic (Welsh,Scottish,Irish)
background for UK caucasians.
Just thought this piece worthy of airing here - Juror
interpretation of CSI type television
http://www.zwire.com/site/news.cfm?
newsid=10059769&BRD=982&PAG=461&dept_id=467992&rfi=6
DNA profiling query
Would anyone care to put on a defence, expert-witness, hat and
give an interpretation of this report from a recent
UK court case ? In particular the possible nature of the
20th allele(s) referred to
in verbatim/transcript quotation marks.
Quote
Forensic scientist F S explained that what is known as SGM Plus testing looks at ten sites on strands of DNA to identify 20 'numbered components' which are compared to an identical test carried out on a suspect's DNA.
If any single component does not match, then the two samples cannot have come from the same person, but if they do match, a statistical analysis is carried out on the odds against it having come from anyone else.
In 1997 she initially carried out the SGM test which was used at the time on four of the internal swabs.
But last year, before W had been arrested, she carried out the more advanced SGM Plus testing.
Of the tests on vaginal swabs, Mrs S told the court: "There were 19 definite components out of 20. The 20th was present, but there was some uncertainty as to whether it was one of two different numbered components."
After W's arrest she tested his DNA, and found the same 19 components plus a possible match to the one she had not been able to identify and there were no differences.
She told T R QC, prosecuting, the chance of it being from someone else was one in one thousand million.
End Quote
I forgot to say this rape and murder happened
7 years prior to the court case.
DNA profile from the defendent on another matter,
6 years later, was the only evidential connection to the murder.
The only other, circumstantial, evidence was that he
lied about ever having been to the crime-scene town.
The prosecution showed he had an operation at the town's
hospital the previous year.
Also his ex-girlfriend originally gave him an alibi
for the relevent night but she later retracted this.
I'm not really sure, since it is a bit out of context. I am sort of
stretching here, but perhaps the 20th allele was at a level below
their interpretational threshold and there was an artifact present at
about the same level. The "20th" allele theoretically could have been
one of the two peaks, but it was not certain which was which.
I really can't tell just from reading this, but this was my first guess.
My concern was if this was a possible interpretation
of the statement as written.
Assuming TWGDAM protocols etc observed and
tests repeated. Also no indication that either sample
was mixed.
Could one sample have given a homozygotic ,say, (18,18) on
one sample and a variant allele (18,18.2) at the same locus
on the other sample.
In other words a 19 point match/20 point mismatch
which of course is a mismatch.
It is possible to have slightly different DNA profiles in different
tissues, due to mutation. This would create a "19 point match and 20
point mismatch." Just because a person has the same DNA profile as
the evidence, it doesn't necessarily mean that the person left the DNA
there (it may be a billion times more likely to find the DNA types if
left by him than another random male, but still a chance). Similarly,
a slight mismatch doesn't mean that the person didn't leave the DNA.
He may still be millions of times more likely to have left the stain.
Basically you would find the probability of finding a person at
random that matches all loci except for the one mismatch. Then you
factor in the odds of a mutation at the mismatched locus. If the
probability of finding a random male in the population with the
matching alleles 1 in 1000 million and the probability of a mutation
at the remaining locus is 1 in 1000, the overall probability would be
about 1 in 1 million.
Taking a more general case.
If we were to take both a blood sample and a
buccal sample, say, from the same individual
at the same time.
Test both samples on know good, calibrated,
equipment is there then a high probability of a
match on 19 + mismatch on one ?
If this is a common occurance or any situation
where there was an anomaly I would expect
tests on different, say 5 loci.
Ignore the anomolous one so match on 18
and it would have to be a match on all
extra 5 loci or a mismatch declared.
Whatever the reason for the anomoly whether
mutation within the given individual
or known systemic failing.
Such as known problem with FGA and SGM Plus, from
Forensic Science International 112 (2000) 151 - 161
sometimes falsely recording say (19,26) as
apparently homozygotic (19,19).
First of all let me state that my numbers were totally made up. I
don't actually know the frequency of a mutation off the top of my
head. I don't know the stats on a mutation like this, I've never
encountered it. Paternity labs encounter this much more frequently.
But the mutation rate is much higher in a given sperm (or egg) cell
than in the totality of the semen.
Lets say for instance that a mutation happens at a given locus in 1
in a thousand sperm cells (numbers made up). Then a child of an
alleged father has a decent chance to have one allele that doesn't
appear to support paternity.
But if the SEMEN from the father is tested, 999 out of 1000 sperm
cells may have the normal type, while 1 may have the mutant type.
This mutation would go UNDETECTED in the semen, since it is swamped
by the normal sperm cells.
Just as a mutation can happen in a given sperm cell, it could happen
in a given blood or epithelial or any other cell. This would cause
maybe one in a thousand cells of these types to have a mutant
allele. This still might not be detected.
But suppose the first cell in the zygote, that would eventually
divide and differentiate into the testes, has a mutation. Then all
of the sperm cells that were created in the testes would show this
mutation. This is much rarer than the mutation for any given sperm
cell. I don't know what this number is, but someone probably has
calculated this. My example of 1 in a thousand is much more common
than this would be. It may be 1 in a million or 1 in a thousand
million, I don't know.
Lets say for example that a mutation like this happens 1 in a million
times:
If the probability of selecting a person at random and finding a
matching profile is 1 in a thousand million, then it is a thousand
times more likely that the DNA would match if it was left by the
DEFENDANT than a random male.
If the probability of selecting a person at random and finding a
matching profile is 1 in a thousand, it is a thousand times more
likely to find the profile if it was left by a RANDOM MALE than if it
was left by the defendant.
So to properly deal with a situation like this, you would need to
take in the rarity of the DNA profile in the population and the
rarity of a mutation event like this. If one is significantly more
rare, it would sway the evidence.
If you simply went to the respective tissue (like obtaining a semen
sample if the difference is in the semen) and it matched there, it
would confirm the mutation and make this whole arguement moot.
I don't know about all labs, but I would guess that they would report
it as a probable match, and give the appropriate statistics. I also
think they would ask for a sample of the appropriate tissue. For
instance if there was a difference between the blood and the semen,
they might ask for a semen sample to confirm the mutation exists in
the semen. It might be difficult to confirm the difference anyway.
It is often easier to confirm that there are slight differences than
to confirm how or why those differences might exist.
The 1 in a 1000 figure seems to be regularly
quoted in the paternity testing area.
This relates to first mitosis at oogenesis where a mutation, of 1 allele,
goes on to produce a mismatch in all the cells of the blood (say),
between parents and offspring. The process of evolution
requires it for survival of succesful variants and suppression
of non-viable variants, for advancement of the species.
There is probably the same process in later mitosis, at the stem-cell
stage, where just one organ or cell-line could vary from
the rest of the body.
Anything later and there would be an infinity
of possible sports, not all the same variant.
There is a minimum of, 200 or so,
cells to pass through the normal PCR process.
An un-explained, just posited, tissue mismatch would
be fine in an academic study but this was presented
in a murder case where the defendent was given a life sentence.
It is absolutely beholden on the prosecution to exactly
explain this 20th mismatch - to the extent of analysing
different tissues of the defendent.
Posing a
prosecution view (accepting the,20 point, billion:1 figure for the moment)
then 19 point match <>350million:1 false match figure
shoots past the 1billion to an off the planet 1000x 350m:1,
if the defendent has a proven mismatch on this 20th allele
between semen and buccal cells.
Taking the defense view (if they had had a DNA expert),
failing to nail down the 20th mismatch, the prosecution
has presented evidence of exclusion,not inclusion.
19 point match + 1 point mismatch is a mismatch -
not a match. They cannot take the best of 20
to declare a match, it is all or nothing.
If I had been in the public gallery of that courtroom
I would have been shouting - foul !
Much thanks for
the input on this very technical point but so far
I've seen nothing to sway me from my original
assessment that this was a highly unsafe conviction.
The other 'evidence' was a very murky night-time CCTV
tape that showed a human form and certainly no
points as to identity.
The defendent lied (understandably) that he had visited,
at any time, this town 25 miles from where he lived. He was
proven to have had a specialist operation at this town's hospital.
His ex-girlfriend changed her testimony, not giving him
an alibi, as to where he was that night 7 years previously.
Misrepresenting statistics - what does it mean?
The most recent information coming out of the HPD lab scandal (see:
http://www.chron.com/content/chronicle/special/03/crimelab/index.html)
is that their forensic scientists have given inaccurate statistics and
insufficient evidence in 23% of the cases reviewed.
It's the inaccurate statistics thing that's bugging me. I've seen
forensic
scientists misrepresenting numbers, overstating the confidence of
evidence,
say match to a jury without explaining what it means and know damn
well the
consequences...
This is also akin to what Joyce Gilchrist of Oklahoma and Arnold
Melnikoff
of Montana and Washington State were doing - lying about the
numbers and in
some cases citing phony statistics. These were not and are not
stupid people
working in isolation from the rest of the forensic community. And
they've
all been doing it for a long time. And they've been promoted and
rewarded
for their work by others.
It's not possible that those at HPD and that Gilchrist and
Melnikoff and the
many others like them are doing this without the knowledge of other
forensic
scientists. Is it?
The first question then is this: Is this malice or ignorance? Do
these
forensic scientists and those who've trained them simply not
understand the
basics of hair and DNA analysis? Is that possible? Because I've
read some
Saferstein I've always assumed that this kind of conduct was
malice... but
that assumption may be misplaced if this problem is so widespread.
The second question is this: what would the results be if similar
independent evidence/ testimony audits were conducted of other
state crime
labs?
Question # 2 should be of concern to every young or student forensic
scientist out there whose working or considering work at a state
police lab.
The lesson of the Huston PD crime lab is that this kind of stuff
only stays
under water for so long. Eventually, it comes to the surface and
then
everyone whose taken a drink from the pool is a suspected carrier.
I would sincerely appreciate any public or private response to these
questions. I'd just like to know what other forensic scientists
think.
They are all doing it not just HPD
If you look for the statistics used as the backbone
behind DNA profiling you find reams of theory
but no simulations or extracted from real data.
I've hunted through all the relevent forensic
science journals and its all theory leading to
the 1 in a billion through 1 in 100 billion
through to trillions and even quadrillions
quoted as fact in court.
I know from my own simulations based on the published
allele frequencies that multi-billion figures are criminally false.
The figure for 10 loci SGM+ as used in the UK
for a 2 million database has to be less than
1 in 1.8 million chance of false matches.
I would say in Gilchrist's case, it was malicious, but I think in
Houston's case it is more of incompetence.
Most of Houston's problems came with mixture interpretation and how
the statistics relating to mixture interpretation are understood and
presented.
Here is a somewhat simplified scenario to illustrate what was going on:
Lets say that at a given locus there are three possible alleles: A, B
and C. Each person has two alleles, so there are six possibilities
(AA, AB, AC, BB, BC, and CC). But AA, BB and CC are only observed as
A, B and C. Let's say the defendant is AA. Lets also say that only
1% of people have AA at this locus. But lets say that the EVIDENCE
sample is a mixture and happens to type as ABC. In this case, you
cannot tell how many people are contributing to the mixture. All six
possible genotypes are included in the mixture. So the defendant is
included in the mixture, but so is EVERYBODY.
Houston reports that the defendant is included in the mixture and that
the probability of selecting a person at random THAT HAS THE SAME
GENOTYPE AS THE DEFENDANT is 1 in 100.
The competent lab reports that the defendant is included in the
mixture and that the probability of selecting a person at random THAT
WOULD BE INCLUDED IN THE MIXTURE is 100 in 100.
It doesn't matter what the probability of having the same genotype as
the defendant, we only care how likely someone would be included in
the EVIDENCE. But these can be very different.
Thanks for responding with this excellent and very clear example.
Given that this is such a basic thing, and that the error is an accurate
display of statistics that are pro-prosecution but unrelated to the specific
circumstances of the evidence in a given case (and a failure to explain that
everyone, not just the defendant, is included as a possibile
contributor) --- is this really just incompetence?
Presenting one set of favorable findings and witholding unfavorable findings
seems incompetent, yes, and more han a little biased, yes. Could it be
intentional but not malicious? Could it be that these paricular criminalists
are so entrenched in the idea that they are "out to get the bad guys" that
they have forgotten that they are supposed to be scientists?
By the way, thanks all for being patient with my questions.
It's just that my experience has taught me that there are very few people
who wake up in the morning and think about how they are going to lie to keep
the bad guys behind bars. Most people who do this kind of thing have simply
convinced themselves that this is good science, or do not have the education
or training to know that it isn't.
Personally I don't know for certain, but I think they were not
intentionally giving statistics for purpose of helping the
prosecution. They certainly may have been, but unlike with Gilchrist,
I have never heard of them saying things like "the prosecutors love me
because they can count on a prosecution..." as she supposedly said.
If they knew what statistics they should be presenting, but presented
other stats, they were certainly malicious. But I sort of think they
just didn't know what stats they were supposed to be presenting.
I thought coursework in statistics and probability was
a standard requirement for any position involving
genetic analysis. I've always been alarmed at the
number of my coworkers who seemingly didn't understand
the limitations of the technology they were using but
unfortunately I can't say that I'm surprised anymore.
In clinical trial situations, use of accession numbers
to identify specimens is standard practice (i.e.:
double blind studies). Is this practised in forensic
laboratories?
Actually you are totally off base. I have seen your models, they show
a clear misunderstanding of the statistics involved.
But I deliberately do not involve statistics.
I have a healthy disrespect for statistics and leave
that nonsense to the Weirs,Budwoles and Baldings
of this world.
I am happy to use published allele frequency
tables as they are derrived from the real world.
On the latest dnas5.htm version, all alleles for
UK caucasians are included which was the previous
criticism.
Otherwise the only straying into statistics is
multi-modal matching and the broadening and sharpening
of tail-offs of the otherwise approximated multi-Gaussian distributions.
This is a consequence from
doing such simulations - unknown otherwise, as I've
found no reference to it in the DNA profile
statistical texts.
Summarised as - the commonest alleles increase
in frequency in matches, rare alleles decrease in frequency ,
and rarest alleles are practically absent from matches. At
least at the level of a 4 million pool.
The square law is known within statistics and I have
an explanation of it and it comes through in these
simulations so no problem there.
Summarised as - if you have x matches in a pool of
y profiles then in 2y profiles you will have x^2 matches.
forensic scientists and dyslexia
i just wondered what the position is with forensic scientists (in
the uk) with dyslexia ie can they still work or do they need a
certificate to state this fact?
thanks
The following errors ( one is architypal
dyslexic w<>m nowhere near each other as keys
on a keyboard ) in one of the
standard forms from the principal UK FSS centre in Birmingham
" The details will only be used, however, for the
for the (sic) prevention and detection of crime "
" consequently some of the following items may
have no data recorded against them if it is not appropriate for the
purpose for
which the record mas(sic) made. "
1 of the 10 loci used in the UK, D8S1179 ,
was printed on this form as D6S502 not
even the correct chromosome.
All undersigned by the then director a
P. E. Cage
-----------
ps No one explained why my use of
statistics was flawed in a post here
a few weeks back.
Test for ex-frozen biological material ?
Are biological stains routinely tested for whether
the serological material has been previously frozen?
Is there a test for previous freezing, if their
is suspicion of such a likelihood?
Hypothetical scenario - habitual rapist goes to area
where prostitutes do their business. He picks up
recently discarded condoms and puts them in his freezer.
Later when committing a rape, using a condom himself,
he deliberately spills the contents of one of his un-frozen
condoms on the victim's clothing.
That's a very clever MO. I have heard that freezing produces crystals in flesh
cells, like muscle, which tend to burst cell walls. I have never heard that it
has done so in semen. I ran across some studies in the freezing of semen in
prize animals (bulls-horses-dogs) that show decreased mobility in frozen-thawed
semen, even with a preserving chemical. The thing is that crime scene semen is
typically collected, dried, then sent to the crime lab, and then processed for
typing. This is usually weeks after collection and the lab might even freeze it
to preserve the stains, especially if case work it backlogged. And, In these
days of DNA testing, where serological testing is almost never done, I doubt if
the physical signs of freezing would be noticed at the lab, in most
circumstances. So unless there was some hint of the MO with the discarded
condoms, I doubt if it would ever come up. Detectives might scope it out, if the
stain recovered resulted in a CODIS hit, and the guy with the matching profile
had an ironclad alibi, like he was in jail, at the time of the crime. Also the
dicks would get suspicious (I hope) if a series of similar rape investigation
resulted in several different DNA profiles, unless he kept re-freezing and
thawing the same condom, which would show increasing physical damage to the
sperm. Kind of a goofy answer, but I hope it helps.
Homozygosity in DNA profiles
Contrary to my usual investigations my recent
has gone counter to expectations.
It is not the case, as generally perceived, that excess
homozygosity would lead to greater likelihood of
false DNA profile matches ( in autochthonous communities
for instance).
In fact the opposite is more likely.
Analysis:
I did 4 off 50,000 runs of 6- loci profiles based on Australian
data.
First run produced 64 off 12 digit matches.
Adding some homozygosity, resorting and checking again
there was only 50 matches.
Another run , without excess HZ, produced 55 matches
Repeating but this time adding the homozygosity at generation,
sorting and match checking, again resulted in only 50 pairs.
Australian caucasian data just for locus D5
Code here / Allele / Allele frequency %
0 8 0.49
1 9 3.08
2 10 6.03
3 11 39.16
4 12 34.36
5 13 15.64
6 14 1.11
7 15 0.12
For unmodified D5 subset of 50,000 profiles
bi-allele totals, firstly homoZ then largest heteroZ counts
00 - 2
11 - 59
22 - 166
33 - 7,647
44 - 5,848
55 - 1,274
66 - 4
Overall excess D5 HZ = 38.7 %
33 excess HZ 43%
44 excess HZ 21.8%
So new 33 count is now nearly the same as the now reduced
34 count and as a consequence is equally likely to influence the
6 loci match situation but only if the excess HZ on other loci
is similar in behaviour.
But if 33 excess HZ was only 20%
then 33 count goes up from 7,647 to about 9,200 but
still less than the 34 count now about 12,400. The square law has
it that the larger counts predominate disproportionately.
Consider 44 situation ( less likely to occur in matches)
but now 6,797 >> 4,482 so critical in comparison.
Now consider the general case.
h is excess homozygosity expressed as scaling factor ie 1.28 rather than 28 per cent.
a is the allele frequency of an alelle that has excess HZ factor of h
and b is the highest (no excess HZ ) allele frequency of the remaining alleles.
Normal HZ is a^2 and b^2
Highest normal hetero-bi-allele frequency = 2*a*b
Excess HZ increases a to (a^2)*h
Highest hetero-bi-allele frequency becomes approximately = 2*b *{a - 0.5*a(h-1)}
= a* b*(3 - h)
So criticality is if a*a*h <> a*b *(3-h)
or a*h <> b* (3-h)
So even if h is high then the b frequency if high
enough in comparison it can over-ride.
Then this effect has to be present in most loci to
affect the overall match probability.
Excess HZ of less than 10 % is generally more likely to reduce the number
of unrelated profile matches in populations of the same type and size
DNA profile error rate now down to 8 per cent
That is 8 in 100 forensic DNA profiles as used in the UK consisting
of 10 markers and 20 datapoints.
Most recent data from journal article
International Journal of Legal Medicine (2004 ) Issue 2 118: p83-89
http://springerlink.metapress.com/app/home/contribution.asp?
wasp=43wrc2wrrncvtvcg9j4k&referrer=parent&backto=issue,4,13;journal,3,
47;linkingpublicationresults,1:101167,1
or http://tinyurl.com/2t484
The GEDNAP blind trial concept part 2.
Latest data for year 2002 which is an improvement
2001, 10 per cent erroneous profiles.
2000 , 14 per cent erroneous profiles.
This data is anonymised so the good and bad are lumped
together - considering the bad do not know they are
bad (or at best, in disagreement with the consensus).
These are the results of GEDNAP ( German DNA Profiling
Group ) blind trials of testing samples at 136 labs in 31
European countries.
First thing to note this was blind trial ( not
double blind ) so all participants knew in advance and
could process immefiately after calibration, not Friday
afternoon processing / interpretation/ transciption etc.
Much of the 'improvement' is because:
Many incorrect results in previous GEDNAP trials had been due to
specific types
of body-fluid stains so those "have since been discontinued because
of
the inconsistency of the amount of DNA present " from the more
recent
trials, despite those sorts of stains being commonly found
forensically.
Then the intractable human error:-
"most of the errors were made each time by a very few number of
laboratories and of course compound errors such as interchange
of two samples caused a disproportionate number of errors relative to
the one mistake made when sampling the wrong test stain. However,
this is not taken into consideration when calculating the error
rate."
( these specific errors included or excluded in the calculated error
rate? )
"the most common type of error has always been transcriptional errors
followed by incorrect interpretation due to failing to recognise
an eror, these types of human error are to some extent unavoidable
under any prevailing circumstances."
I repeat this is results from labs knowing they were
being tested and presumably on their best behaviour.
Then the more technical/ systemic problems producing mistyping
because of
stutter, 'long alleles', unrecognised rare alleles etc.
19 correct out of 20 datapoints is fine in 8 out of 100
DNA profiles if you are identifying body parts, say,
from a small pool of possibles
but not for the usual forensic purposes of implied uniqueness.
I'm at the moment processing 10 million simulated
profiles generated to CODIS 13 loci form and N. American
Caucasian allele frequencies and all profiles
generated totally randomly.
Whether 10 million is enough to show at least one
false match I don't know. Indicative info is that
because of the rather ridiculous TPOX allele 8 in CODIS
(2 chances of 54% for each person to inherit allele 8 )
means for these purposes CODIS is little more than 12 loci
in effect. Australian NIFS 9 loci simulation gives a baseline
population of about 400,000 ( on dnas9.htm URL below )
and UK , FSS 10 loci simulation
gives 1.1 million (file dnas5.htm on URL below).
CODIS simulation result will be file dnas11.htm , probably
next weekend , ( flavour of what is entailed it requires
about 1.5 GB of free disk space for processing ).
DNA Evidence
I have seen this brought up a number of times which leaves me with 1
question can a person be convicted in the UK on DNA alone ?
In Canada a person may be detained and questioned with the possibly of being
arrested based on the presence of DNA, but for a conviction you would need
more than just the presence of DNA.
As with any Database complete uniqueness seldom exists, for example in 1998
in Lethbridge Alberta a store was robbed and the clerk was shot. A camera
from a neighboring bank filmed a 1967 - 1969 (The camera didn't catch the
headlights so they couldn't be more accurate) Mustang leaving the scene. The
vehicle had an Alberta license plate. There were 7 vehicles matching that
description. I had a 1968 mustang registered at the time I was picked up for
questioning after a few RCMP questions I was given a ride back to work and
never heard from them again. So I don't see why complete uniqueness is such
a big issue.
This is how I understand it. 4 people enter a building the criminal left
some DNA behind. Out of those 4 people 1 persons DNA matches the sample at
the scene the other 3 don't. So the DNA would exclude 3 people.
What is truth ?
This article in one of today's UK papers
http://www.guardian.co.uk/life/farout/story/0,13028,1225015,00.html
Recounts an experiment first done in the 50s
but still relevant.
Forensic scientists can act as sheep or be deluded
or just act under peer pressure like anyone else.
Why else would they believe what forensic statisticians
say with only theoretical backing about probabilities
of false matches with DNA profiles. ?
Still outrageous figures eg 640,000,000,000 in Oz this week,
are quoted with absolutely no empirical foundation.
You could not get away with it in any other 'scientific' discipline.
As the authorities in charge of these large DNA databases
are not divulging how many unrelated matches are contained
within the arrestee side then I had to go out and simulate
one myself to find out. As far as I can tell no-one else
has done this ( or at least published the results )
Details of a full allele 2 million 'profile' computer
simulation
http://www.nutteing2.freeservers.com/dnas5.htm
It is my understanding that this group is a professional group
related to the use of science in matters of law. From the posts I
have read it appears many of the listees are students of forensic
science. As forensic scientists, your duty does involve debate.
However, this debate must be based upon scientific principles, not
emotion, loyalty or politics. I am extremely disappointed in the
latest political thread. As forensic practitioners your
RESPONSIBILTY is truth. Truth based upon the physical and
unemotional analysis of evidence. The fact that someone saw a
photograph does not constitute the analysis of evidence. More
importantly, the input of faith or political affiliation should
never
play a part in your unbiased observations or opinions. If voicing
your political, religious or other non-scientific views is your
interpretation of forensic science, then you have failed as a
forensic scientist.
Which planet breeds forensic serologists ?
http://www.presstelegram.com/Stories/0,1413,204~21474~2300828,00.html
Article Published: Wednesday, July 28, 2004 - 8:35:45 PM PST
Quote
When analyzing DNA, forensics experts tend to talk about numbers
rarely heard outside college classrooms and science laboratories.
They use terms like quintillions (a number with 18 zeros),
sextillions (21 zeros) and septillions (24 zeros).
And when it comes to matching DNA profiles, each zero matters. In the
case of Mark Wayne Rathbun - the alleged Belmont Shore rapist -
forensic serologist Thomas Fedor has calculated a chance of one in
844 septillion that someone other than Rathbun left his DNA on the
left breast of victim Jane Doe No. 2 in May 1998.
But how did he get that number?
Such complex calculations, Fedor said, are based primarily on a
population survey that indicates how many people in a group are
likely to have the same information stored in their genetic markers.
For instance, Fedor said, if one in five people have the same
information stored in one specific marker, then there's a 5‚percent
chance that two given people share that marker. If one in 10 people
have another specific marker, then the percentage of people who share
both these markers is calculated by taking 10 percent of 5 percent.
The calculation is simply repeated for each of the markers studied,
Fedor said.
"If you take 10 percent of 10 percent of 10 percent of 10 percent 13
times, you're coming up with a very small number," he said.
This formula, Fedor said, is called the Product Rule and is widely
accepted in DNA analysis.
Old Cases Involving Faulty DNA Testing Targeted - Texas
It does not require corruption - just blind faith in a very
flakey technology, only 8 per cent of 10 loci DNA profiles
are correct , Source: International Journal of Legal
Medicine (2004 ) 118: p83-89. That is crime-scene
or arrestee profiles.
And slavish following of forensic statisticians.
Throughout the 1990s they were spouting, even in court ,
false match probability of 1 in 37 million for 6 loci profiles.
That was until admission of a crime-scene profile
matching 2 separate unrelated arrestee profiles in the (UK) NDNAD
(R v. Watters appeal) and case of Raymond Easton
forced a halt to that dangerous nonsense.
The real 6 loci false match data for a population of
4 million is between 94,000 and 380,000 matches.
So they upped it to 10 loci which certainly improves things
but not the multi-billion figures now spouted.
Again for a 4 million population the false match figure
is between 5 and 32 pairs. 5 is unrealistically low (
parthonogenic - no coancestry ) and 32 too high as all
having similar ancestry as myself - one parent and his
parents from one English county and mother and her parents
from another English county ).
Because the technology is so flakey it is necessary to
consider 19 out of 20 matches as a match not a mis-match.
This is argued permissable because a mutation on any locus/
allele is possible (0.01 to 0.3 percent on any of the 20 loci/alleles)
As long as further exploration to confirm this
conjectured mutation - but do they ?
19 out of 20 if identifying identity of a decomposed body, profile
already
known, or someone named as a suspect in an enquiry but not
where the suspect has been implicated from a DNA
database trawl.
By 19 out of 20 I mean any 9 pairs matching and one allele match
between the remaining
pairs.
For the completely artificial situation for totally random
profiles ( no co-ancestry at all ) then for any 19 in 20 then about
120 matches in 4 million or 1 in 33,000.
For the opposite situation and excessive co-ancestry then
28,000 free 19 from 20 matches or 1 in 140
of a 4 million population.
Until real-life data becomes available for co-ancestry
( parents & grandparents from same or different country,
from same or different county etc )
tabulated against allele frequency minima then these are the
outer bounds as it stands.
They only arrest
the first person who happens to match in a DNA database,
they don't wait for ten years, say, for a third match
to come along.
Relevant simulations and results are on dnas5.htm
and dnas6.htm on URL below. I would like to be
able to point to academics or forensic scientists
who have done such simulations but I seem to be the
only one to determine false-match likelihoods from
real-world allele frequencies. Afterall it does
not require any fancy high-power computer
The USA with 13 loci CODIS profiles cannot be smug as one
of there loci/alleles is TPOX (8) which 79 percent of people
have one or more. Also increase the number of loci
and you increase the chance of an error in any one
locus/allele so throwing out the whole profile,
falsely implicating or falsely eliminating.
DNA in the Dock
A Codis, 13 loci , 26 number unrelated false match.
What about all the others without such a cast-iron alibi ?
http://www.suntimes.com/output/news/cst-nws-dna01.html
Quote
DNA links crime to woman with alibi -- she was in jail
November 1, 2004
BY FRANK MAIN AND ANNIE SWEENEY Crime Reporters Advertisement
At first, police thought blood taken from the scene of a North Side
burglary solved the crime because of a DNA match linked to a woman's
genetic
profile.
But it turned out the woman had a solid alibi: She was in prison at
the time
of the break-in about two years ago, authorities said.
Now, police are working with state officials to see whether there was
a
breakdown in the steps leading to the woman's DNA profile being
entered into
the database, known as CODIS for Combined DNA Index System.
Investigators are checking whether the state prison system mistakenly
put
the woman's name on another inmate's DNA sample. The woman submitted
a mouth
swab to state prison officials in May when she was paroled on a drug
conviction, sources said.
Investigators also are looking into the possibility of the Illinois
State
Police crime laboratory entering the wrong information into CODIS.
There are other possible explanations, as well.
Robert Hovey, supervisor of the DNA Review Unit of the Cook County
state's
attorney's office, said investigators must make sure the blood sample
was
directly related to the burglary and that the woman was never inside
the
apartment. He cautioned against a rush to any kind of judgment about
the
system.
"We don't know if the bloodstain is related to the burglary," said
Hovey,
who did not know the details of the case. "But DNA is only going to
prove
presence. It is not necessarily going to prove someone committed a
crime."
Lincoln Hampton, a State Police spokesman, said the state lab was
looking
into the case.
"They are reviewing that case and will present their findings to the
Chicago
Police," he said.
The burglary in the 1300 block of West Eddy was among a pattern of
about 70
break-ins that police have been investigating.
The case intrigued Kathleen Zellner, a Naperville attorney whose work
led to
the exoneration of four men in the 1986 rape and killing of medical
student
Lori Roscetti. Zellner sought DNA tests that excluded the men as
sources of
semen on Roscetti. Two other men were charged.
"DNA has been the gold standard of evidence," Zellner said. "If you're
implicated, you're sunk, and if you're exonerated, you're home free.
But
there are bound to be mistakes. I don't know of anything quite like
this."
End Quote
30 year-old DNA
I keep reading of cold case prosecutions from recovered
DNA profile evidence from 20 or 30 year-old evidence.
I've searched FSI and IJLM for validation checks studies on
the long-term stability of DNA and nothing found.
These old samples are usually stored at ambient temperatures,
not frozen.
Frozen is the choice for current samples specifically for DNA
analysis so obviously a preferred state. Degradation
at room temperature leading to reduced number of loci
being copied/identified but is there increased stutter
or mis-interpretation of those loci that are identifiable.
Anyone know of any validation experiments in this area ?
Mixed sample DNA profile analysis - art or science ?
or artifice.
I'm looking at a 2003 published journal report
concerning analysis of a mixed sample. Unusually
suspect as major components and victim as minor.
Suspect was an employee and presumably prosecuted
for rape.
Analysis with Powerflex profiling so 15loci + 1 .
Electropherograms of mixed stains, and referenece
samples from suspect and victim are shown.
With standard threshold setting of 150 units shows 11
alleles consistent with the victim and one vWA (14)
that is alien to suspect and victim.
Only 2 persons concerned in this sample, going
by the maximum number of any allleles.
Unmistakable peak well above threshold level shown
in the plot but not apparently labelled as 14 - (tippexed out
before publication ? )
There should ideally be 18 alleles from the victim so
they lowered the threshold to 50 units and this brought up
to 18. This plot is not shown in the publication , surprise ,
surprise. The >150 plot shows traces of the missing 7
but also 2 other alleles that must be >50 but alien to
victim and suspect.
A medical anomoly (trisomy ) concerning the victim
was explained but nothing about this vWA (14) or the
other unwanted 2 as spontaneous mutation or anything.
A low resolution pdf of this journal article is on
http://www.cmj.hr/2003/4403/18DROBNIC.pdf
Croatian Medical Journal 44(3) :p150 - 154, 2003
The British Library photocopy of the printed version of this article
is slighly clearer but not much. The highly questionable
spikes are evident in both PDF and printed versions.
On page 351 on the bottom row , section B,
two blocks of spikes from left, are spikes labeled 15,16 and 17
but to the left of them is definitely an unlabelled spike at position
14.
The following will need magnifying the pdf.
The 3 lines of this B plot , including arrowed sites are
D3, THO1, D21, D18, Penta E
D5, D13, D7, D16, CSF, Penta D
XY, vWA, D8, TPOX, FGA
The extra spikes greater than the 50 threshold that are not arrowed
would be D21 (29) [ third set on top row ]
and D5 (10) [ first set of middle row ]
and probably numerous others.
There is at least one unlablelled spike in the SGM
'A' plot of D19 (13) first in bottom row but all
this 'A' plot are indistinct in this plot, again exclusionary
This vWA(14) on its own in a just world should have
been exculpatory let alone the other alleles that had to
pulled out of the mush - the reason for threshold limits
as far as I am aware.
As silence so far and exploring other such cases I get
the impression this behaviour of cover-up of such
anomalies to be the norm in this area of forensic 'science'.
Can anyone give some examples of the use of mathematics in forensic?
What kind of mathematics used? Is it calculus? Probability? Etc?
Regards,
For anyone confused by this followup , it refers to
journal article
http://www.cmj.hr/2003/4403/18DROBNIC.pdf
and thread (archived on NNY) from mid Nov , 2004
http://archive.nnytech.net/sgroup/FORENSIC-SCIENCE/3321/
http://archive.nnytech.net/sgroup/FORENSIC-SCIENCE/3401/
I take your point but in this article IMHO there is
the most blatent of litteral cover-ups.
The machinery & software automatically labels all
alleles of peaks above a certain threshold.
The peak '14' definitely appeared in the plot but
the label has been deliberately removed and no
reference to it in the text either.
I thought, until getting involved with forensic scientists,
above all else scientists explore anomalies - it is
absolutely fundamental to the scientific ethic.
The trisomy was addressed for example.
My more general point about trying to pull
incriminating data out of the mush below the
threshold level has not been addressed. I have
since obtained a photocopy of the printed version
of this article and my observations are even clearer
in the printed version than the low res pdf.
Let me break the unbearable "cover-up" silence.
First, have you ever heard of stutter? The vWA loci commonly
experiences this artifact of amplification. Hence what you consider
a 14 allele is probably a stutter peak. Notice I said probably. I
would not be so foolish as to make any kind of definite assertion
(such as yourself) based on the limited data presented in said
article. Without the raw data for analysis no reasonable conclusion
can be drawn from the Figure 1B plot.
Second, let's assume your assumption of a 14 allele at vWA is
correct for the moment. The presence of the extra allele in a mixed
stain is not exculpatory. This is elementary. Again, assuming the
14 allel was a legitimate peak, its presence would only indicate a
possible 3rd contibutor to the stain. Extra alleles do not exclude
contibutors from a mixed stain.
Sorry to destroy the shooter on the grassy knoll... no conspiracy
here!
Re: vWA
For forensic purposes ( false inclusion and false exclusion)
vWA is highly suspect for the reason you state.
Doubly so, because it is the worst for false homozygosities
see FSI 143 (2004) 47-52
vWA and from this journal article D8S1179,
FGA, D5S818 and D18S51 should be removed from
multiplex analyses for forensic purposes, far too
high occurance of false homozygosities.
Only picked up like this cmj case study by splitting
samples and running on different systems, primers etc.
Let me break the unbearable "cover-up" silence.
First, have you ever heard of stutter? The vWA loci commonly
experiences this artifact of amplification. Hence what you
consider
If you really believe your statement about "proper" science not requiring
interpretation, then you have little true understanding of the scientific
process and its application.
_ALL_ scientific data requires interpretation by a human scientist in order
to draw conclusions about it. Computers and analytical instruments only
produce data, they can't interpret it. They are idiot savants - great
number crunchers but unable to interpret and apply their products. They
can't recognize relationships and draw conclusions on their own. They only
know and understand relationships that human beings define for them. Until
artificial intelligence is perfected to the point where it can duplicate the
synthesis of new information and recognition of subtle connections between
seemingly unrelated data that human minds routinely accomplish, the GIGO
(Garbage In, Garbage Out) principle applies. Computers can suggest
inclusions or eliminations, but only human experts can differentiate those
suggestions and confirm which is correct.
No result produced by a scientific instrument or process should be accepted
without scrutiny and evaluation by a human expert. Anyone who would trust a
machine to draw scientific conclusions for them is a fool destined for
disaster. It takes a human being with appropriate expertise to decide
whether or not the data that instruments and computers produce is even
valid, let alone to determine how to apply and interpret it. Your
"dispassionate filters" are figments of imagination that will never have the
necessary intelligence and insight to do what you seek, unless cybernetics
can one day produce a true artificial mind with a consciousness all its own.
We are still quite far from turning such science fiction into reality.
Regarding your last comment with its slanderous insinuation, please listen
up. Reputable forensic scientists, like all reputable scientists, work for
one thing and one thing only: the discovery, documentation, and revelation
of scientific facts. We don't care what direction those facts lead; we will
accurately characterize and report them regardless of who they help or hurt.
We don't work "for" the police, the prosecution, _or_ the defense, and our
conclusions aren't influenced by who pays our salaries or "commissions" our
work; nor are our careers influenced by whether or not our results are what
our "customers" were looking for (our careers are influenced only by the
quality of our work). We "tell it like it is," and let the chips fall where
they may as it is no concern of ours. We are not partisans within the
adversarial system, and have no stake in how cases are ultimately
adjudicated.
We are simply fact-finders with no goal or end to achieve other than the
presentation of those facts to the investigators, attorneys, judges and
juries; it is THEIR jobs to determine the case outcome, not ours. Any among
us who allow themselves to become prejudicially influenced by ANY part of
the criminal or civil justice system only disgrace themselves and are justly
pilloried and expelled from our profession.
The goals and desires of legal advocates are irrelevant and of no
consequence to us or our work, other than to advise us of the questions that
need to be answered in a given case. Any police investigator, prosecutor or
defense attorney who tried to influence our work product or our report
content would be bluntly told to go pound sand. None has ever dared to
suggest such a thing to me more than once, I can assure you. Objectivity
and impartiality are the very basis of this or any science, and we of the
forensic science profession guard ours particularly zealously.
Are we clear now?
Thank you for taking timeout to read the report and reply in full
and on theme rather than discussing extraneous matters.
vWA (14) is a perfectly valid naturally occuring allele ,
it can also result erroneously in a plot, from stutter.
By what automatic process, on one pass through the system
and one set of primers,
can recognize the signature of a stutter as distinct from
a real allele. Then when the post analysis ladder-check
binning and labelling software cuts in why is the anomolous
peak not labelled 'S' ,say, for stutter.
Concerning pulling data out of mush.
Would you agree with me there are further minor peaks
above the re-scaled (lowered) threshold that are in fact
higher than the peaks arrowed in the 'mush-field' ?
I can assure you those peaks are in the printed
version and are not an artifice of a low resolution internet
PDF file version.
http://www.cmj.hr/2003/4403/18DROBNIC.pdf
Figure 1
The 3 lines of this B plot , including the arrowed sites are for
D3, THO1, D21, D18, Penta E
D5, D13, D7, D16, CSF, Penta D
XY, vWA, D8, TPOX, FGA
The extra spikes greater than the 50 threshold that are not arrowed
would be D21 (29) [ third set on top row ]
D18 (14) ( Fourth set on top row ]
and D5 (10) [ first set of middle row ]
Neither suspect nor victim have these 3 alleles
Note the software could not be relied upon to output
a reconfigured plot with lowered threshold level
without labelling at least the 3 peaks D21(29) D18(14) and D5(10)
hence, IMHO, the standard normal level threshold plot
with non-standard added arrows. Not even fudging it
by choosing some threshold level between 75 and 50 could things
be manipulated to include the 'wanted' alleles and exclude
the 'unwanted' alleles.
A fundamental tenet of science is the principle
of reproducability.
If I pick up a calculator and accurately enter a request
for the tangent of 40 degrees and get the figure
0.839099631 assuming accurately transcribed.
I expect you in another country to do the same
on a completely different calculator and get the
same figure with maybe a rounding diference on last digit.
In this CMJ report a sample from same source was
put through 2 related processes and ended up with
2 different results concerning vWA. The interpretation that can
be made with any authority is that vWA cannot be relied
on and should not be a part of a system that incriminates.
This 'science' is not reproducable so cannot be called
a science.
I posit that there must be 10s of thousands of erroneous
profiles in databses because of vWA stutter and 10s of
thousands more because of false homozygosity recorded as
real. Similarly but a third to half, the vWA erroneous profiles rate,
for false homozygosity recorded as real in D8 , FGA and D18
and to a lesser extent D5
Data from FSI 143 (2004) 47-52
No doubt a 14 vWA allel can be real; however, you can not, I repeat,
you CAN NOT determine the position of that peak or any other
unlabeled peak from the data in that article.
The number of primer pairs is irrelevant with respect to stutter. If
you believe the number of primer pairs has anything to do with
stutter, then you clearly do not know what stutter is. Also I
personally have not seen software that labels stutter. Further
more, software does not interpret stutter, a trained analyst does.
Finally, no I do not agree there are minor peaks that are valid
alleles after lowering the cutoff threshold. At least not with the
information provided in the article. To make any true determination
I would (and so would you) require the original data and appropriate
software.
Have you actually ever done any DNA analysis? If so, what kit and
instrument? I would never dare to make the claims you are making
without seeing the data for myself. To do so (IMHO) would be
irresponsible and wreckless. Every analyst I know takes the
interpretation of STR data very seriously. We do not want to send
innocent people to prison by forcing data to fit the story. This is
a serious business we are in.
If it was all a proper science then there would
be no need for 'interpretation' .
The analysis should be automatic and unambiguous.
Where there are anaomalies or ambiguities they should
be picked up by dispassionate system filters and checks not
post-analysis human 'interpretation' in your parlance
and sleight-of-hand in my parlance.
Bear in mind the whole process is in the first instance ,
and usually the last instance, commissioned at the
behest of the prosecution.
Perhaps we are talking at cross-purposes.
On the second page , page 351, of article on
http://www.cmj.hr/2003/4403/18DROBNIC.pdf
Figure 1 B and on the bottom line , first group is
the Amel one and the next group from the
left is the vWA plot, as far as I know.
Are you telling me you do not see a peak to the left
of the 3 peaks labelled 15,16 and 17 ?
Agreed I cannot state with any conviction that this
peak, I'll call vWA(*), represents an allele 14 or 13 or some variant
as only a graphical plot. But there is definitely something
there projecting well out of the mush-field, to my reading of it.
Without any prior knowledge ( ie victim and suspect only
show alleles 15,16,17 ) and one pass using the
PowerPlex system, how can the analyst declare this vWA(*)
peak is an artifice of the system rather than real.
In normal processing a validation
repeat on the same sample, ie Powerplex here, is likely to show
the same anomalous / real allele .
Agreed, processing with another kit, SGM in this case,
highlights this fourth vWA peak as questionable but
my point is in normal, routine processing, what would
indicate that this fourth peak was a stutter artifice. ?
What signature (eg shape or position ) in the original electrophoretogram
in isolation from any other information, would show ,
even if just presumptively rather than conclusively,
that this vWA(*) was not real ?
I would conjecture that an SGM run was done because
the vWA (*) did not fit victim or suspect, ie prior 'knowledge',
that run failed to bring the minor contributions out of the
mush-field, so nothing gained there.
<>
So you expect the data to analyze itself? Any science requires
interpretation - how do you think new drugs are made? The researchers first
have trials in which they plot data from experiments and then analyze or
interpret what the data means for that drug's usefullness. Even a simple
GC/MS of a drug needs to be interpreted to make sure everything is accounted
for. Handing raw data to someone untrained and unfamiliar with the process
would be the equivalent of handing me a soldering iron and expecting me to
solder a microchip in place or handing my little brother a scalpel and
expecting him to remove someone's appendix. Data must be
analyzed/interpreted by a trained professional, just as a MD analyzes blood
work and/or any test performed to find out what ails you.
One of the definitions of science is "knowledge gained by experience"
(American Heritage Dictionary) or "knowledge obtained from the systematic
study of the structure and behaviour of the physical world, especially by
observing, measuring and experimenting, and the development of theories to
describe the results of these activities" (Cambridge Dictionary).
Instruments and/or computers do not gain knowledge by experience. They only
do what they are programmed to do. Everything to a computer is reduced to
1s and 0s. A human programmer compiles a list of steps the computer must
perform in its interpretation. So even if we allow the computer/instrument
to perform the interpretation we are in fact still having a human perform
the interpretation. However, the human involved is usually a computer
science professional that does not necessarily have training in genetic
analysis. Therefore, allowing the computer/instrument to perform the
interpretation would possibly/probably cause greater errors than having a
trained human analyst review the data and interpret the computers
interpretation by rules set up during training and experience.
Thank you for so clearly expanding on what I said in my most recent
(still not posted) reply to Nonarevers. His view of science is very
idealistic and unrealistic. I almost used the GC/MS example in my
reply:)
"If it was all a proper science then there would be no need for
'interpretation' ."
So you expect the data to analyze itself? Any science requires
interpretation - how do you think new drugs are made? The
researchers first
Automatic and unambiguous? First a lesson on software. Who writes
software and filters? People do! So while it may (someday) be
possible to write filters to fit every scenario, there is still the
fact the filters are determined by human beings. As such, STR
analysis is subject to interpretation since there will always be a
human factor somewhere in the process. I disagree with the word
ambiguous. It implies DNA analysis is uncertain or that there is
always two ways to view the results. This is simply not true. When
we encounter hard to interpret data then we always err on the side of
being very conservative when stating an inclusion. Sometimes we
can't make a determination and state as such.
As to your prosecution remark. Clearly it implies that all DNA
analyst are just "cops in white lab coats" who merely manipulate the
data to "make" the prosecutions case. While I can't argue there have
been embarassing incidents within the community; a few bad apples do
not reflect the attitude of the community as a whole. I know that
within my lab we all would love to be independent of law enforcement
just to remove even the appearance of favoritism. However, who
writes my paycheck is completely independent of the work I do.
Frankly I love to see a good case result in a conviction. I also am
equally happy to see a suspect exonorated by my work. It all pays
the same!
A final remark... I have to this day not seen any "proper science"
that is automatic and unambiguous. The uncertain nature of things is
why we have science. Automation... a luxury.
What is the major, unreported, flaw in this sort of statement seen in
court houses around the world, reported in last day.
http://www.thereporter.com/Stories/0,1413,295~30195~2543067,00.html
<...>
At that time, state Department of Justice criminalist Nicola Shea
stated that Wilson's genetic profile would be expected to occur in 1
of 96 billion Caucasians, 1 of 180 billion Hispanics and 1 of 340
billion African-Americans. The victim's genetic profile, Shea said,
would be expected to occur in 1 of 110 trillion Hispanics, 1 out of
140 trillion Caucasians and 1 out of 610 trillion African-Americans.
Shea noted that the population of the world is estimated at 6.5
billion to 7 billion.
<...>
I doubt no more than 100 million profiles have been determined
in the whole world and no more than 20 million using one
compatible profiling system.
Courts are for statements of fact not unfounded
and untested speculation like the
above quote. The correct statement should be of the form
number of unrelated matches in the known 20 million profiles
, or whatever number, and then from the rarity/commonness of allele
frequencies
concerning this defendent's alleles how likely he is to
have a false match.
The UK government will not disclose how many unrelated
matches there are in the 2.7 million arrestee side of the NDNAD
and the FBI remove any from data before handing over to academe.
From doing multi-million profile simulations I know
there are so many unrelated matches after 100 million
population/ database that quoting billions let alone
trillions ...... heptillions is tantamount to perjury.
DNA from cremated ashes ?
http://www.channelnewsasia.com/stories/afp_asiapacific/view/121429/1/.
html
<...>Tokyo announced Wednesday that DNA tests did not match ashes
purported to belong to Megumi Yokota, who was kidnapped in 1977 when
she was 13 by North Korean agents looking to teach spies about Japan.
Chief Cabinet Secretary Hiroyuki Hosoda said that another set of
cremated remains brought from North Korea by a Japanese team last
month were not those of another victim, Kaoru Matsuki, as claimed.
"We have told (Matsuki's) family members that the remains displayed
someone else's DNA," the top Japanese government spokesman told a
press conference. <...>
Which planet breeds forensic serologists ?
http://www.presstelegram.com/Stories/0,1413,204~21474~2300828,00.html
Article Published: Wednesday, July 28, 2004 - 8:35:45 PM PST
Quote
When analyzing DNA, forensics experts tend to talk about numbers
rarely heard outside college classrooms and science laboratories.
They use terms like quintillions (a number with 18 zeros),
sextillions (21 zeros) and septillions (24 zeros).
And when it comes to matching DNA profiles, each zero matters. In
the
case of Mark Wayne Rathbun - the alleged Belmont Shore rapist -
forensic serologist Thomas Fedor has calculated a chance of one in
844 septillion that someone other than Rathbun left his DNA on the
left breast of victim Jane Doe No. 2 in May 1998.
But how did he get that number?
Such complex calculations, Fedor said, are based primarily on a
population survey that indicates how many people in a group are
likely to have the same information stored in their genetic
markers.
For instance, Fedor said, if one in five people have the same
information stored in one specific marker, then there's a 5,percent
chance that two given people share that marker. If one in 10 people
have another specific marker, then the percentage of people who
share
both these markers is calculated by taking 10 percent of 5 percent.
The calculation is simply repeated for each of the markers studied,
Fedor said.
"If you take 10 percent of 10 percent of 10 percent of 10 percent
13
times, you're coming up with a very small number," he said.
This formula, Fedor said, is called the Product Rule and is widely
accepted in DNA analysis.
There's nothing wrong with the product rule, as it is a well
established
statistical principle. However, there are those who will tell you
that it
is too simplistic an approach to use in DNA analysis because
genetic markers
are not necessarily all independent variables. There are other more
complicated statistical calculations which may apply and may or may
not
still produce astronomical results. Others will state that the
question of
"what is the chance of two people having the same marker?" isn't the
appropriate question, but rather "what is the chance, given the
frequency of
this marker in the relevant population, that someone other than the
suspect
could have deposited the stain?" or rather "what is the chance that
two
people who share this marker could have been at the scene?" - The
answers to
these questions require a different approach, involving more
advanced
concepts such as calculation of "prior odds" and "posterior odds".
I'll let someone more qualified than me in this area (I'm just a
drug
chemist) explain the details of the above, but there is one glaring
error in
the below quote I can point out. A "1 in 5" chance is a 20%
chance, not a
"5% chance" as stated below, so the example calculation begins with
a faulty
figure. I would wager that Thomas Fedor was misquoted in this
regard, as
I'm sure he can do primary-school level mathematics, but I'm less
sanguine
about the abilities of journalists who write articles like this
one. One
should never trust the average journalist with math, let alone with
science
- they often mangle both into unrecognizable forms. If you seek
accurate
information in the mass media (itself not a particularly reliable
source)
regarding science, try to read pieces written by journalists who
specialize
in science news and who themselves have a background in science,
such that
they actually have some understanding of their subject matter.
Most of the
rest know nothing of which they speak, and almost invariably get it
wrong
(or at least not entirely right) when reporting about it.
No, - statistics are valid for extrapolating from
billions or trillions dowwn to say a country's
population. Not the other way around let alone
a small subset of a country.
Please tell me where all these billions or
trillions of profiles have been determined.
I've still not found anyone else in any discipline
who has taken realworld allele frequencies
and simulated multi-million DNA profile database
detailed fully on
http://www.nutteing2.50megs.com/dnas5.htm
If , like me, a defendent should have in all (UK NDNAD terms)
20 alleles, characterised by allele frequencies
greater than 8 per cent then talk
of billions or trillions before an unrelated match is
found is totally unproven and as such should not
be stated in courts.
Another disturbing aspect appeared this week
http://www.harktheherald.com/modules.php?
op=modload&name=News&file=article&sid=43190&mode=thread&order=0&thold=
0
or
http://tinyurl.com/4r2r5
This prisoner has no matching MO and no other
evidence just a profile match which IMHO is
entirely inadequte when databases are containing
millions of profiles now.
I still have not found out which planet has billions,
trillions, quadrillions .... heptillions of humanoids to
take profiles of to be able to scientifically and legally make
such statements in an earth-based court.
Once you get to a database of a few hunreds of millions
in the Australian NIFS system, a million or so in the UK
NDNAD system or 10 or so millions in the CODIS database
then you have unrelated false matches appearing.
Remember CODIS, for false match purposes, is effectively
only 12 loci, not 13 because 79 percent of people have at least
one TPOX (8) allele.
All you can state properly in court is there is not a match
to this defendent in this database of ? number profiles. There are
?? unrelated matches within this database. Then it is
a matter of stating how the individual alleles in this defendant's
profile compare with the allele frequencies of occurance in his
genetic/ racial supopulation.
Quoting totally unsubstantiated multi-billion numbers
are just there to colour the perceptions of the jury -
they have no scientific validation. The hundreds of
papers by forensic statisticians are a blot on
an otherwise admirable molecular biological science.
They have woven Scotch mist to make something
admissible in court.
It seemed a very strange comment "I have a healthy disrespect for
statistics" from a person who is using statistics to the extent of
conducting statistical simulations in attempting to prove a point!!
Chris Anderson
It is very easy to understand this stance on statistics. If you (or
one of your close family members) was accused/arrested/convicted of a
crime, you also might try to start some movement to falsely discredit
the science that caused the accusation/arrest/conviction. It is very
easy to avoid the science and math and create your own world where the
truth doesn't make sense. Fortunately, people like "nonarevers" will
never be on a jury here in America, for at least two reasons:
1) He is totally biased towards his own brand of nonsensical pseudoscience
2) He isn't an american
I'm not really sure who he is attempting to persuade with his crusade
of idiocy, but it is probably best for all of us if we just ignore his
rants.
I'm not understanding this stance on statistics. Here in KY our
database
seldom generates hits because the markers we use are highly variable and
highly individual. We use 13 STRs. I've seen the research and read the
papers. I've seen the research and statistics applied to horses, dogs,
cows, cats and mice. It makes sense. The article given may not
make sense,
but then again how many mainstream media types can actually get
something
like that right? DNA has been through the hearings and been
approved. Its
not perfect, nothing is, but its a definitely better than IEF blood
typing
for generating more unique results. At least that's my opinion and
that of
my coworkers.
Well Adam, his (or her) posts seem very emotional to me. I would guess that
"nonarevers" has suffered personal harm because of what he/she perceives as
undue reliance on DNA evidence in a particular case. Perhaps a lover or a
relative convicted by it. I have had a few testy exchanges with him/her. But I
actually think that the "statistics" are so overwhelming, at this point, that a
positive result is simply a match, and the numbers don't count for very much.
"one in the world's population" pretty much says it all. There will be times
when there are false results due to poor procedure or wrongdoing but, in the
normal course of things, its as routine and boring as a blood alcohol or drug
analysis.
A true forensic scientist would only pontificate
after experimenting to confirm or counter some
hypothesis.
Where is your evidence that you will only
find false matches when databases get to
off-the-planet 'populations' of trillions ..... heptillions ?
I have the authority of being able to speak as
someone who has simulated multi-million
profile numbers based on real-world data
not some hypothetical head-in-the-clouds
theoretical statistics.
Nonarevers,
Circular logic will not get those people released from prison. DNA is a
scientifically well founded way of showing that two samples share a common
source. The stats are so overwhelmingly supportive of the techniques used that
even if your thesis about false matches were reasonable, the odds against more
than one of these false positives would be astronomical. The argument has been
had and won by the science. Defense lawyers lost this one. The FBI just calls
it a match, as I understand it. Concentrate on the integrity of the laboratory
procedure, as a procedure, and on scene contamination. That is a much more
likely source of reasonable doubt. Your not doing these people any favors by
pinning their hopes on air, or on some esoteric philosophy of science argument.
You have a better chance of convincing juries that martians did it, a much
better chance, they believe in martians.
No one is lying or exaggerating. There is no vast conspiracy to falsely imprison
your friends. There are mistakes made, even wrongdoing on the part of a very
few forensic workers. That does not change the fact that DNA profiling does what
it claims to do. We believe it because we have no choise, its scientifically and
legally valid. I'm not trying to shut you up, I'm really not, but I dispute the
practical validity of you arguments. Happy New Year!
Nonarevers.
I would like you to do a simple favor for me. Take 26 equally
weighted dice. Throw them together and see if all of them land with 3
on their top face. It should only take you a few seconds. Now grab
another 26 dice and do the same. Keep repeatinig with sets of 26 dice
until you have all 3's up. It will brobably only take you a couple of
trillion years or so. Oh, you won't live that long? You don't have
3.4 quintillion dice? There aren't even a 3.4 quintillion dice in the
whole world? Then I guess statistics and probability theory must not
be valid on our planet. Just because something has a large (or small)
number doesn't mean that the theory behind it is invalid.
When you take many independent observations (whether they are tosses
of dice or combinations of independantly inherited genes) and try to
figure out the probability of finding any specific combination of
those events, the numbers can be very large if the number of
observations are large.
I've given you a free pass for most of the year, but your drivel has
finally irked me to a response. I am now satiated and will allow you
to pollute the web with misinformation until you instigate a response.
I find it amazing that you don't even appreciate
the fundamentals.
Your dice are proper balanced dice but real
life uses loaded dice.
Just taking as an example a THO1 'UK Caucasian' dice
of '8 sides' they are
not equal probability of 0.125 of each face turning uppermost
For THO1
'Face5 ' probability 0.002
Face6 , 0.241
Face7 , 0.194
Face8 , 0.108
Face8.3 , 0.001
Face9 , 0.14
Face9.3 , 0.304
Face 10 , 0.012
So faces 9.3 and 6 are far more likely to
appear than 1/8 at each toss. Repeat this twice and
for each of the other loci and there are always
more commonly occuring alleles.
For people like myself with no criminal record
but my DNA profile permanently on record and
allele frequencies at all loci greater than 8 per cent
then I am far more in the firing line to be falsely
arrested in the future just because my profile
matches a crime-scene profile.
I have done this 'dice tossing' with all loci/alleles
represented and the results do not tally with
your assumptions.
At least I have simulated on computer all these
dice, modelled on real-life allele frequency data (loaded dice).
For 6 loci , 12 datapoints, then unrelated matches start
apearing at about 10,000 throws. Remember court experts
,were stating at that time, pre-2000 false match probabilies
of 1 in 37 million.
For 9 loci ,matches start appearing about 200,000
For 10 loci about 1 million
For 13 loci about 10 million
Which planet breeds forensic serologists ?
Here is a ragbag collection of statements in various courts.
I thought, niaively, that courts were for presenting facts
and proveable/ testable evidence.
"The odds against another individual matching Danielle's profile are
in the quadrillions, said Mitchell M. Holland "
"Chidambaram said. ``With this new method, the chance is one in
quintillions So the chance of identifying a particular person is
pretty high""
"... occurring in approximately 1 in 21 sextillion of the Caucasian
population, 1 in 650 quadrillion of the African American population,
1 in 420 sextillion of the Hispanic population."
"Thomas Fedor has calculated a chance of one in 844 septillion that
someone other than Rathbun left his DNA... "
Well Adam, his (or her) posts seem very emotional to me. I would
guess that "nonarevers" has suffered personal harm because of what
he/she perceives as undue reliance on DNA evidence in a particular
case. Perhaps a lover or a relative convicted by it. I have had a few
testy exchanges with him/her. But I actually think that
the "statistics" are so overwhelming, at this point, that a positive
result is simply a match, and the numbers don't count for very
much. "one in the world's population" pretty much says it all.
There will be times when there are false results due to poor
procedure or wrongdoing but, in the normal course of things, its as
routine and boring as a blood alcohol or drug analysis.
Wally Lind relates, "Concentrate on the integrity of
the laboratory procedure, as a procedure, and on scene
contamination."
Oh Wally, he has. A quick read of his website shows
that Paul (nonarevers) has a suggestion for the
potential rapists out there...
"For rapists all you have to do is pick up recently
discarded condoms from prostitute working areas. Place
in your freezer and leave the unfrozen contents of one
smeared on your next victim - use a condom yourself of
course. Forensic scientists do not routinely check for
previous freezing of such samples. They often place
such SoCo exhibits in a freezer anyway."
http://www.nutteing2.50megs.com/dnapr.htm
I stopped responding to his posts a long time ago and
treat all messages posted by him as spam.
Just my two-cents.
Perhaps I should of acknowledged assistance on the background
technical
matter by reference to message 6367 on 05 March ,2004 of this group
"The thing is that crime scene semen is
typically collected, dried, then sent to the crime lab, and then
processed
for typing. This is usually weeks after
collection and the lab might even freeze it to preserve the stains,
especially if case work it backlogged. And, In these days of DNA
testing,
where serological testing is almost never done, I doubt if the
physical
signs of freezing would be notice
d at the lab, in most circumstances. "
INTERNET CRACKPOT HOGWASH, straight from the anti-establishment lunatic fringe.
A guy who gives forensic advice to rapists on his website, is in need of a few
"fundamentals" himself. Like maybe some common decency.
I thought this subject was beaten to death in
Summer/Fall of 2003. Message #5174 put a stop to him
for a while but I suppose he's found renewed energy as
of late...
nonarevers,
So in other words you don't believe in extrapolating statistics? I don't
claim to be a statistician, but from my limited statistics experience what
DNA analysts say is scientifically valid. Forensic scientists are generally
very cautious when saying something matches. It seems to me that you are
taking only one side without throughly knowing the science of both sides.
As far as I know, there are no other planets with humanoids - much less
trillions or quadrillions. However, science here on Earth has led
scientists, statisticians, and forensic analysts to believe that the
probability of someone having the same DNA profile is very large. Have you
looked into any of the databases to see how many, if any profiles are the
same? Have you examined numerous pieces of evidence and done the procedures
necessary? Do you know your human genetics and how random assortment and
crossing over and other genetic phenomenon occur? Or are you going on an
oppositional rampage insisting that because you don't understand and/or you
don't want to understand that the science is unfair and untrue? I've looked
at the link you gave, and all I found was someone's idea of the truth that
they spout with some knowledge, but not understanding. I'm very hesitant to
accept things just because someone says its true - especially when that
someone is going against all other proof. The person responsible for that
website you keep linking to is more than a little off-base. He/she
indicates in this area
"The second line of text on the above covering letter ending "the record
mas made". THIS IS A DYSLEXIC ERROR - letter reversal/inversion - not a
typing error, w and m are nowhere near on a keyboard (also occurs in
dyslexics concerning letters u and n, d and b etc). The other indicator for
dyslexia is this erroneous initial letter is the same as the initial letter
of the next word. Typing letter q, e, a, s or d instead of w is just sloppy
fingers on the keyboard a letter m indicates a sloppy mind. If this person
is involved with the likes of labelling DNA samples or batches then god help
us. "for the for the" shows lack of proof-reading / lack of managerial
supervision.
What do defence attourneys make of such evidence of incompetence when it
arrives on their desks from this country's supposed leading national
forensic laboratory."
Not only is defamatory to any number of dyslexics out there (of which I am
one), it is also highly prejudiced. Also, while the writer goes on about
'sloppy minds' and 'sloppy fingers' they do not puncuate correctly nor do
they spell correctly (defence to my knowledge is not a word - defense is the
proper spelling). To call a dyslexic a 'sloppy mind' is like calling
someone blind ignorant. Also, even if the person inputting the data is
dyslexic and turns a word around or inputs an improper character, peer
review will catch errors. That is why at least in my state every profile
entered into the system has at least 3 analysts review it - first the
initial analyst reviews their data, they then submit it to another qualified
DNA analyst for review. This analyst reviews not only the procedures
followed and statements made, they also check the spelling (for lack of a
better term) of the alleles. After that analyst is finished reviewing the
data, a DNA database analyst reviews the work all over again, making certain
that the information printed out by the instrument is what appears in the
analysts report. Often, another analyst will review the data at a later
time during an audit or similar process. We are human, so errors may be
made, but it would be statistically unlikely that 3+ people would all make
the same mistakes. I am proud to be trained in molecular biology as well as
trained in forensic serology. I admit that erros do occur, but the entire
group cannot be judged on the basis of so few. It would be like saying,
since we know some people steal/cheat/kill/etc the entire human population
is untrustworthy and wrong at all times. Quoting only small sections of a
literature reference is like quoting only a word out of a sentence. I don't
know what has happened to you in the past to make you see things so jaded,
but I hope for you the best.
Dyslexia is a potential problem in many areas of work
eg delivery driver, postman etc. But it pales into insignificance
compared to being employed in forensic science even only as
a clerical officer producing reports. A mistake by a
dyslexic in such an environment could lead to the death
of an innocent in many USA states.
Leaking roof over forensic lab sample-preparation
area (Texas) is considered OK by one set of lab
directors so perhaps dyslexia is perfectly acceptable
also. Not in my books though.
I've only been in receipt of one communication
(ignoring repeat copy due to administrative incompetence )
from the supposed premier UK forensic lab
in Birmingham, England. This letter was mimeograph signed by the
then director a Dr P E Cage. In it was prima facie
evidence of dyslexia and lack of proof reading by whoever
scripted the letter and also evidence of technical incompetence
concerning confusion over D6/D8 loci.
The then director put his name and signature to this
document so condoning the lack of ability of his staff
by lacking the ability to spot the errors, himself, or equally
wrongly by not reading what he puts his signature to.
Comments about spelling or use of English should
be on the appropriate board/group.
To exonerate someone I would expect on average
at least 5 'alien' peaks in an electrophoretogram
that are not present in the incarcaree or victim +
incarceree, mixed crime-scene stain. One mismatch
can be (suspiciously too often ) explained away as
spontaneous mutation but 5 or so ?
Unless you are somehow saying that a historic criminal
deliberately seeded a crime-scene with someone else's DNA
on the off-chance that a future technology would exonerate him.
In theory it takes (USA) 26 in 26 numbers to, truely or falsely,
implicate a person but only one in 26 to exonerate.
I wonder if any forensic lab technicians have been
falsely dismissed because it is easier to dismiss one
lowly employee rather than overhaul an ingrained
country-wide system. People dismissed because it is
implied they cross-contaminated samples whereas in my
'theory' the results could well be correct. But as it is
deemed impossible for false matches then blame
the worker and not the system which can bungle
along untrammelled producing yet more
miscarriages of justice. What the hell, the vast majority
being implicated via DNA databases are criminals
anyway. My details are in such a database, not because
I have a criminal record or conviction , but because of
the activities of corrupt social workers and then
a Stalinist regime in political power in the UK.
Followup to previous concerning the Minimum Allele
Frequency Formula before matches appear.
Scaling the AF tables for the high AF situation and
then using the minimum number for matches formula
produced the following.
Calculated minimum numbers for sub-sets of profiles like
mine of > 8 per cent AFs , also for >3 per cent ,Codis case.
6 loci UK Caucasian for >8% then 2,400
10 loci UK Caucasian for >8% then 280,000
9 loci Australian Caucasian for >8% then 44,000
9 loci Australian Aborigine for >8% then 21,400
13 loci USA Caucasian for >8% then 2 million
13 loci USA Caucasian for >3% then 10.2 million
If anyone here would like to compare their own DNA
profile through the published AF tables for the
relevent principal population subgroup and reveal
what their minimum AF (FST/theta - like) is I would be interested.
My minimum is 8.7 per cent, for UK Caucasian , all
the remaining 19 are higher than 8.7.
Michael,
Bravo - even I the nonstatistician understood your logic. Of course, that
implies that for someone to understand they must at least acknowledge
logic... I'm not sure Paul goes that far. Very good presentation of his
thesis only more thought out and logical... without the 'fundamentalist'
mentality.
I would dearly love to interrogate the likes of the FSS NDNAD .
The last time anyone published any figures of matches
within this database was around 1997 ie 6 loci days.
From journal Forensic Science International 95 (1998) p30.
Concerning data in the UK DNA database as of 04 October 1996
when there were only 6311 samples from the London
area and 573 from the Cardiff area.
"A small number of unresolved duplicate pairs of
profiles were present in the regional data :10 pairs
within the London region and 1 pair in Cardiff. "
Apparently not investigated further, for aliases etc or at least reported.
About the same time there was a TV report of
about 600 matches (transcript/provenance of this not available
other than details of the program scheduling ) for the whole NDNAD
in mid 1990s.
Despite repeated written parliamentary questions
the Home Office/ FSS will not divulge the numbers of
10 loci matches. Until they do I rely on my
simulations. Standard conventional statistics does
not explain the numerous occurances of false matches,
and I'm not refering to contamination events.
It was not just dyslexic errors.
In the printout of my profile D8S1179 on the 8th chromosome
is referred to as D6S502 on the 6th chromosome. To me
showing gross technical incompetence.
An exact copy of this recorded delivery letter was sent recorded
again a few days later. Strongly suggesting managerial
incompetence. All I can say it is fortunate that the UK
no longer has the death sentence.
I am a scientist , not prose writer. Full of poor grammar
etc, I'm sure, but I will only correct the English if it
leads to ambiguity.
As far as multiplying your figures together I
end up with the same order of result.
BUT
Firstly DNA profiles consist of directed number pairs
and 2pq is usually greater than p^2 where p is the
largest and q the next largest (q usually < 2p) . Intuitively I
thought excess homozygosity would lead to
greater false match likelihood - not true. This is yet
another myth that seems to be prevalent.
Secondly I don't have the computer power to
simulate 10s of millions of CODIS profiles.
I can really only vouch for 10 loci, multi-million
'profiles' simulation. In all the runs I have done then any
10 loci matches have occured where the minimum allele
frequency in those profiles is >= 5 per cent.
Again for the UK, in the low millions on the NDNAD, it
is people like me >8% AF that are in the firing line.
I suspect when the numbers get to 10 million odd
then >3 %ers start getting snagged.
Incredibly there seems to be no awareness within
the forensic community of this phenomenon. I would have
thought it would have been an automatic adjunct where a
suspect is found by database match Just flashing up
the minimum AF of this person relative to their ethnicity to
avoid the Raymond Eastopn / Peter Hamkin false arrests let
alone the others that have gone on to prosecution.
Conventional statistics is
not showing up these "all-high allele frequency matches". People with
a single <1% AF occurance in their profile are likely
to be fairly immune from false matches. It requires
someone with greater computer power and background
research data to simulate for unknown half-brothers and
cross-linking of loci/alleles where these persons may
start to be implicated.
I have been very rigorous checking that any matches
are not a manifest of the pseudo random number
generator and they are not.
I do not consider myself to be exceptional as far as
genetic ancestry. One parent and his parents from
one county and other parent and her parents from another
county. Coincidence that my minimum AF is 8% ? - I don't
know. There just is not anything in the public domain
correlating minimum allele frequencies to ancestral/ichthonous
backgrounds. For all I know all 6 forebears from the same
county may mean a high probability of such people having
min AF of >=10%.
I would like to be able to point to other researchers
who have explored such avenues but I seem to be
on my own. I have a healthy disrespect for statistics.
I've gone back 5 years, in effect, and compared the
standard statistical analysis with my simulation
method but for just 6 loci as the computer processing
is easily handlable by anyone. It puts it all
into stark contrast.
UK 6 loci AFs for Caucasians (allele in brackets)
Locus / p / q / 2pq
vWA 0.27 (17) 0.219 (18) 0.118
THO1 0.304 (9.3 ) 0.241 (6) 0.147
D8 0.333 (13) 0.209 (14) 0.139
FGA 0.187 (21) 0.165 (22) 0.062
D21 0.258 (30) 0.226 (29) 0.117
D18 0.164 (14) 0.145 (15) 0.048
Product/reciprocal gives near enough 1.2 million
Then using my random simulation generation Visual Basic code
for full 10 loci, employing all IJLM published alleles.
Doing a cut-down run of 45,000 'profiles' then
finding matches on the first 6 loci and ignoring the
last 4 loci gave these 12 pairs of matches.
Actual results of a run on 07 Jan 2005
vWA,THO1,D8,FGA,D21,D18
(17,17)(6,7)(13,13)(21,22)(28,30)(14,15)
(17,17)(6,7)(13,13)(22,22)(28,31.2)(12,13)
(17,17)(7,8)(13,13)(23,23)(29,32.2)(15,16)
(17,18)(6,9)(13,14)(20,23)(29,30)(18,18)
(17,18)(7,9.3)(13,14)(21,23)(30,31)(12,12)
(17,19)(9,9.3)(13,15)(21,25)(29,30)(14,17)
(16,16)(7,9.3)(9,13)(23,23)(29,30)(14,18)
(16,17)(6,9.3)(13,14)(22,23)(28,30)(14,15)
(16,19)(7,8)(13,13)(18,20)(30,32)(13,14)
(16,19)(7,9)(10,12)(22,24)(30,31)(16,18)
(14,16)(6,8)(13,13)(19,20)(28,29)(14,16)
(14,17)(7,9.3)(12,14)(22,24)(30,31)(12,14)
The minimum AF in the above is for D8(9) of 0.013.
There is no correlation between 1.2 million
and 12 in 45,000. There is no point in discussing
10 or 13 loci before fathoming this glaring anomaly.
The VB code and macros are detailed below.
This is just plain text , so anyone can see there
is nothing like virus code - it is compiled at Run-time.
For longer runs I have automated the procedures
a lot but a 45,000 run is short enough not to
require more than single sorts per sub-file.
Requirements: a copy of Word 97 with Visual Basic
and Macro handling (in Tools section ).
I only have an 8 year old PC
and don't know if the code works unamended on later versions.
Copy the generator routine beween Sub and End Sub
in the Run/ Create section of Word / VB.
Set to 45,000 (actually 44999 ) and find/replace all dec14 file names
to jan08 (or whatever) and Run - takes about 1 minute on my pc.
Gives 8 non-empty string subfiles, convert back to standard
representation at the end.
Sort each one in turn using Word text sort - takes about 5
minutes in total.
Copy and paste back into one file - about 4 minutes in total.
Copy the "find matching pairs " routine to VB and
change to 45000 and file name and run - about 2 seconds.
My run produced 12 '6 loci' matches ,
changing to 13 gave 2 off (6 loci +1) matches, 14 and no 7 loci matches.
Because a random process anyone repeating this process
instead of 12 matches you may end up with only 5 or maybe 20
or anything but something of order 10.
To give standard representation use the final
conversion routine , set to ? (12 in my case)
and correct file names then in output file find/replace
,"", to )( to tidy up and delete the final 4 pairs in
each profile.
This is base-line totally random matching, no factoring
in of co-ancestry, cross-linking etc.
That's interesting since I went to your website and
observed that YOU found no 20 digit (10 STR marker)
matches in a run of 795,224 profiles using your macro;
found one 10 marker match in 546,224 profiles; another
one in 10 marker match in 1,220,712 profiles; and yet
another one in 10 marker match in 999635 profiles.
That's a total of 3 matches in over 3.5 million
profiles for all 10 STR markers using your simulation.
You failed to run more than 10 loci, so I would be
cautious about drawing conclusions about 13 loci since
your own data doesn't support this supposition.
It also appears to me that you have a biased method of
limiting the number of alleles to (generally) ten.
For example, you ignore all of the larger FGA alleles.
Granted, these are rare, but your simulation doesn't
reflect authentic, real world observations (e.g. there
have been 51 observed alleles at D18, you used 16
alleles in your simulation).
I courteously acknowledge that you looked into it.
I can only use the published data. I had initially
only used a maximum of 10 alleles (0 to 9 in the strings) and
was advised to add the minor ones and revised the
structure to cope (0 to 9 and a to z ) but it made no
observable difference as far as matches were concerned.
False matches in 10 loci only occured in the 18 I've
observed at AF > 5 percent. For 6 loci > 1 per cent.
These are the 18 'profile' matches found in numerous
multi-million runs,
I have done one 10 million 13 loci run and no match found but
I don't have the computer power to go beyond that.
But the immediate problem is how to
correlate a figure of 1.2 million by one method with 12 in 45,000
in the more basic 6 loci case. 12 in 45,000 is more in
agreement with published matches
eg the Janet Chaseling (Australia) one 7 loci (14 point ) match in
9 loci from a study of 5,500 unrelated people. That is any 7
from 9 , not the first 7 ,say.
And the other reported matches in the 6 loci NDNAD
when the total was less than 840,000.
Statistics versus Simulation.
To-a-man forensic scientists seem to use the product rule
for determining what the likelihood is of finding a matching
DNA profile to a given defendent.
The Product Rule: - Mathematical rule that the frequency of occurrence of
several independent events is
equal to the product of their individual frequencies.
I have done exact simulations of large DNA profile databases
and the figures in no way tally. Would anyone care to ponder?
Glossary
Locus a recognisable site on a specific chromosomal DNA that has a repeated
code sequence of bases A,G,C or T .
Alllele is the number of repeats of this code at that locus
Allele frequency (AF) is the fraction of a population having
that particular allele at that locus, determined from sampling
real people..
So in table below the maximum 'p' is 27 per cent
for locus vWA and allele 17. The next most frequent 'q'
is 21.9 percent. 2pq because there are 2 alleles at each locus
because one is from mother and one from father. Which, from
which, is usually not known and so expressed as directed
pairs (smaller,larger) by convention.
UK 6 loci AFs for Caucasians (allele in brackets)
Locus / p / q / 2pq
vWA 0.27 (17) 0.219 (18) 0.118
THO1 0.304 (9.3 ) 0.241 (6) 0.147
D8 0.333 (13) 0.209 (14) 0.139
FGA 0.187 (21) 0.165 (22) 0.062
D21 0.258 (30) 0.226 (29) 0.117
D18 0.164 (14) 0.145 (15) 0.048
Product/reciprocal of 2 pq column gives near enough 1.2 million for most
commonly occuring pairs, ie worst case situation.
Forensic scientists expect the first random matches
to start occuring with a population or database greater
than 1.2 million. For 10 loci they state billions and
for USA , 13 loci, then trillions.
Random simulation generation Visual Basic code
for full 10 loci, employing all IJLM published allele frequencies
, URL below. I had to throw out 2 pseudo-random
generators which weren't up to the task.
Doing a cut-down run of 45,000 'profiles' then
finding matches on the first 6 loci ( ignoring the
last 4 loci as set for 10 loci ) gave these 12 pairs of matches,
ie 24 'profiles'.
Actual results of a run on 07 Jan 2005
for 6 loci vWA,THO1,D8,FGA,D21,D18
(17,17)(6,7)(13,13)(21,22)(28,30)(14,15)
(17,17)(6,7)(13,13)(22,22)(28,31.2)(12,13)
(17,17)(7,8)(13,13)(23,23)(29,32.2)(15,16)
(17,18)(6,9)(13,14)(20,23)(29,30)(18,18)
(17,18)(7,9.3)(13,14)(21,23)(30,31)(12,12)
(17,19)(9,9.3)(13,15)(21,25)(29,30)(14,17)
(16,16)(7,9.3)(9,13)(23,23)(29,30)(14,18)
(16,17)(6,9.3)(13,14)(22,23)(28,30)(14,15)
(16,19)(7,8)(13,13)(18,20)(30,32)(13,14)
(16,19)(7,9)(10,12)(22,24)(30,31)(16,18)
(14,16)(6,8)(13,13)(19,20)(28,29)(14,16)
(14,17)(7,9.3)(12,14)(22,24)(30,31)(12,14)
(ignoring the last 4 pairs in each generated profile)
No way involving just the most frequently
occuring pairs of alleles.
There is no correlation between 1.2 million
and 12 in 45,000. 12 in 45,000 seems closer to
the reported situation . A 7 locus match in
5,500 unrelated Australian people and
300 6 loci pair matches in the UK forensic DNA database
when the number of samples was about 200,000.
The problem would seem
to be related to the "birthday problem" ie
number of people for 2 people in the group to
share the same birthday and the number for
2 to share the same specified birthday.
The reported matches in the vicinity of 6 loci are
'Janet Chaseling' 7 loci match in 5,500
Forensic Science International 95 (1998) p30 - all 6 loci
10 pairs within 6311 of the London region, arrested and sampled
and 1 pair in 573 samples from Cardiff region, arrested and sampled
And the TV reported 300 pairs , 6 loci, when the NDNAD
had about 200,000 arrestee profiles in about 1996/7.
I spent about an hour doing a 200,000 run using the
semi-automated procedure (URL below)
"split sorting" hands-on the '0' ,'34', '35' and '45' subfiles,
the remainder 10 to 79 sorted by macro.
This generated 291 matches on 6 loci
43 on 6 loci plus first of next pair
7 pairs of 7 loci matches.
There is a square law for all this " Double the
population and you quadruple the number of matched
pairs approximately". Some co-workers derrived this law -
it's a first order simplification because triples, quadruples matches etc
soon start confusing the situation.
Previous simulation - 12 pairs in 45,000 I could predict
when scaling from 45,000 to 200,000
(200,000/45,000)^2 x12 or 237
This square law effect means that once you start
getting matches , whether 6,9,10 or 13 then it becomes
a rapidly accelerating situation above that. For example
in an otherwise 10 loci database soon after the fist
paired matches appear then triple matches on 18 or even
19 datapoints rapidly become manifest.
The simulations as used here are totally random
with nothing factored in to represent shared genetic history which
of course increases the number of matches.
It is as though they were all parthenogenic , spontaneously
born, with no mother or father. Just the allele frequencies
of all 45,000 or 200,000 are in agreement with the
reported background population AFs.
Match numbers in "ball-park" agreement with the reported situation
and nothing like 1 in 1.2 million chance of a 6 loci
pair match.
Neither UK nor USA have published how many higher
number loci matches are within their databases. The FBI
removes any before passing to academe and the FSS
keep a tight hold.
It is as though you are setting up a lottery
draw with balls of different sizes and masses
and expecting the results to agrree with the
normal balanced ball situation.
Reductio ad absurdum.
Decided to do a 15,000 run - the maximum number
that can be sorted in one go by my copy of Word97.
Resulted in 2 pairs of matches on 6 loci ,
ignore the 7 to 10 loci. As this is a random process
the next run may be only 1 or maybe 3 or even 0
matches, although unlikely 0.
"15565624263618261335"
"44165536135723451746"
Started as undirected pairs
"441655361357A1435345"
"44615563317523457146"
and
"51566542266318261335"
"51565642623611057545"
Converts to standard form
vWa,THO1,D8,FGA,D21,D18,D2,D16,D19,D3
(14,18)(9,9.3)(13,14)(20,22)(29,31.2)(14,17)(17,24)(10,14)(13,14)(15,17)
(17,17)(6,9.3)(13,13)(21,24)(28,30)(16,18)(18,19)(12,13)(13,16)(16,18)
Again ignore loci 7 to 10
URL below . For this seriously cut-down case reduce the
10 subfile statements at the end of the generator
routine to just one and no If / End If.
Change filenames to suit and change run number to 15000
2 pairs in 15,000; 12 pairs in 45,000; 291 pairs in 200,000.
Where is your justification for 1 pair in 1.2 million
for 6 loci false matches ? Again no co-ancestry
(high minimum allele frequencies ) or cross-linking
factored in, all plain parthenogenic.
Another USA false DNA profile match
A very simple explanation although extremely awkward
so not mentioned in this newspaper report.
DNA profiles (twins apart) are not unique.
DNA findings questioned in 1969 slaying
http://www.mlive.com/news/kzgazette/index.ssf?/base/news-12/1105788089296710.xml
Part Quote
Saturday, January 15, 2005
ANN ARBOR -- Gary Leiterman's DNA was identified on five locations of the pantyhose Jane Mixer was wearing when her body was found in 1969, a Michigan State Police crime lab analyst said Friday.
The chances of someone else having the same DNA profile is more than 100 trillion to one, the analyst said at the Gobles man's preliminary examination.
A drop of blood found on the back of her hand, however, was earlier identified by the same analyst as belonging to a convicted murderer who was only 4 years old when Mixer was killed.
<...>
Someone else who is getting his come-upppance for mis-applying
the product rule - Prof Sir Roy Meadow. He of Meadow's law,
One cot death is unfortunate, two is suspicious and three is murder.
He manafged to correlate probabilities of 2 in 3 and 1 in 1 with
73 million to one.
see
http://www.mth.kcl.ac.uk/~streater/cotdeaths.html
Bull,
The Analyst screwed up the procedure. That is the simple explanation, and the
true one.
A very simple explanation although extremely awkward
so not mentioned in this newspaper report.
DNA profiles, even USA ones, (twins apart) are not unique.
DNA findings questioned in 1969 slaying
http://www.mlive.com/news/kzgazette/index.ssf?/base/news-
12/1105788089296710.xml
Part Quote
Saturday, January 15, 2005
ANN ARBOR -- Gary Leiterman's DNA was identified on five
locations of the pantyhose Jane Mixer was
wearing when her body was found in 1969, a
Michigan State Police crime lab analyst said Friday.
The chances of someone else having the same DNA profile
is more than 100 trillion to one, the analyst said at
the Gobles man's preliminary examination.
A drop of blood found on the back of her hand, however, was earlier
identified by the same analyst as
belonging to a convicted murderer who was only 4 years old when Mixer
was killed.
<...>
I come from a scientific background where if theory
does not fit the facts then you have to revise the theory -
not ignore the problem and hope it disappears. Just because
the likes of Weir, Evett, Balding and Gill use the product law
and hundreds of publications mention it - it is unproven.
Repeating unproven material , however often repeated,
does not make it true.
Since DNA typing was invented in 1984 in England, 16 years AFTER "nonarevers"
case, there is something very very fishy about his post, don't you think?
The DNA profiling was done in 2002 on evidence preserved (presumably in
police custody) since 1969. The crux of the controversy appears to be that
one of two men identified in a mixed sample taken from the victim's clothing
would have been only 4 years old at the time of the crime.
Hello all-
I have been reading posts for a short while now and can't help to feel
disgust with this topic that keeps appearing from "nonarevers". This case
he is refering to happened in 1969! Do you not think that as time has
advanced, so has technology? Not only that, but for goodness sake, people
make mistakes. I am not saying it is excusable but since you keep bringing
up USAs errors, has your country made any mistaken DNA IDs? Im sure they
are far more advanced than those scientists here. [hidden sarcasm] Do you
not have anything better to do?
That would seem to be a problem, even if it was pointed out by our listmate.
What a strange turn around - you blaming one of
your colleagues and me blaming the system where
the background mathematics of false matches is erroneously
applied. No its not billions or trillions before false
matches start occuring.
Results first
For completely random
'parthenogenic' profiles , no parents and no
ancestral history factored in.
Then estimated starts for the effects of
co-ancestry.
Results in agreement with published results for 6 loci
and the remainder agreeing with simulations.
6 loci pre-2000 FSS, UK Caucasian: 9,900 so starting at about 2,000
10 loci FSS, UK Caucasian: 2.3 million so starting at about 500,000
9 loci NIFS, Australian Caucasian: 420,000 so starting about 100,-000
9 loci NIFS, Australian Aborigine: 290,000 so starting about 70,000
13 loci CODIS, USA Caucasian: 20.5 million so starting at about 5 milion
There is still
the square law where doubling of the number of profiles
quadruples the number of matches.
To go any further, to bring into the real world,
I need minimum allele frequency to ancestral background
correlation data, which is not in the public domain AFAIK.
Unknown half-brothers (wrong side of the blanket) reduce
the number N even more as also any cross-linking of
loci and allleles.
General formula for determining the minimum numbers
of totally random profiles before the first false match occurs.
For n loci 1..... 6 (9,10,13,15 or any number)
and m (valid) alleles at each locus and 2 per locus.
So Allele Frequencies are AF1 ..... AFm
Let Sn be the sum of the squares of AFs at locus n
ie Sn = AF1^2 + AF2^2 +...... + AFm^2 for each n
Let Qn = Sn^2 for each n
Let p = (Q1 * Q2 * .... * Qn ) [(2-S1) * (2-S2) * .... * (2-Sn)]
Then N = minimum number before finding a match is
N = SQRT (2/p)
For 6 loci results using UK Caucasian AF data
Locus Sn Qn (2-Sn)
vWA 0.1938 0.0376 1.8062
THO1 0.2195 0.0482 1.7805
D8 0.1974 0.039 1.8026
FGA 0.1341 0.018 1.8659
D21 0.1673 0.028 1.8327
D18 0.1236 0.0153 1.8764
so p = 5.45 * 10^-10 * 37.2 = 2.027 *10^-8
and N <> 9,900 for 6 loci
For 10 loci the extra factors are
Locus Sn Qn 2-Sn
D2 0.1213 0.0147 1.8787
D16 0.2218 0.0492 1.7782
D19 0.2379 0.0566 1.7621
D3 0.2068 0.0428 1.7932
So new p' = p * (2-S1)(2-S2)(2-S3)(2-S4) * (Q7 x Q8 x Q9 x Q10 )
p' = 2.027 * 10^-8 * 10.56 * 1.752 * 10^-6 = 3.75 * 10^-13
and 10 locus N <> 2.31 million
Simulations gave about 1.8 million
Australian 9 loci , for Capital Territory Caucasian
Locus Sn Qn 2-Sn
D3 .2056 .04227 1.7944
vWA .1861 .03463 1.8139
D5 .3006 .09036 1.6994
D8 .1971 .03885 1.8029
D18 .1205 .01452 1.8795
FGA .1419 .02014 1.8581
D21 .1585 .02512 1.8415
D13 .217 .04709 1.783
D7 .1763 .03108 1.8237
p = 5.525 * 10^-14 * 208.5
= 1.152 * 10^-11
so N <> 420,000
USA 13 loci Caucasian using RCMP data
Locus Sn Qn 2-Sn
What about those that have used DNA profiles to be exonerated? Are we
freeing people falsely according to your theory? Where do you draw the
line?
Certainly you would have to agree that even if what you say is true - it DNA
analysis is much better than eye-witness accounts and circumstantial
evidence!?
In a perfect world, no one would get falsely accused, nor would any crimes
exist - most of us would be out of a job. Life would also probably be very
boring and innovations would not be made because the need would not be
there.
Followup on Ruelas case
This is all getting ridiculous.
Yet again quote of trillions.
The figure for CODIS 13 loci before false matches occur
is 20.5 million, totally random profiles with absolutely no
co-ancestry. Factor in the fact everyone has ancestry
etc brings that figure down substantially.
http://www.mlive.com/news/jacitpat/index.ssf?/base/news-
11/110615430071040.xml
DNA test results still a mystery
But chances that DNA found on murder victim aren't convicted killer's
are astronomical
Wednesday, January 19, 2005
By Steven Hepker
Staff Writer
Some skeptics doubt a report this week that authorities found the
blood of a convicted killer from Jackson on a 1969 murder victim --
when John Ruelas was 4 years old.
But Washtenaw County investigators say blood on the left hand of Jane
Mixer, a University of Michigan law student whose body was found in a
cemetery, was that of Ruelas.
"I don't think it is a mistake. I think it is his blood," Washtenaw
County Assistant Prosecutor Steven Hiller said. "The chances it
wasn't him are astronomical."
A state police DNA analyst said the chance of a random match is 1
trillion to one, Hiller said.
Officials followed up the database match with fresh DNA samples from
Ruelas with the same results, Hiller said.
"It is odd and a circumstance we want to explain," he said.
The Washtenaw County murder was decades old before DNA evidence
surfaced in criminal cases, but police had preserved evidence from
the victim, Hiller said.
State police interviewed Ruelas in prison regarding the DNA match.
They also have interviewed the man charged with killing the UM
student, 62-year-old Gary Leiterman of Gobles. There is no known
connection between the two men.
Hiller said Leiterman's DNA was found on the victim's pantyhose.
He declined to discuss the results of interviews with the two men,
except to say, "we are still looking."
Ruelas, 40, is serving 20 to 40 years in prison for second-degree
murder in the Jan. 25, 2002, beating death of his mother, Margaret.
He had routinely beaten his mother over the years, culminating in
what investigators alleged was a brutal beating in their St. Clair
Avenue home.
Her face and head were pounded purple, and she suffered 11 broken
ribs, making it nearly impossible to breathe, police said.
The Ruelas family was living in Detroit in 1969, Hiller said. The
father died in the 1970s. John and Margaret Ruelas moved into a
duplex on St. Clair in late 2000 or 2001, neighbors said at the time
of the murder.
They previously lived in Battle Creek. He married a woman in 1996 and
they divorced in 2000. His ex-wife sought three different personal
protection orders against him in Battle Creek, according to court
records.
Police said by the time John and Margaret Ruelas moved here, he had
battered his mother for years. In 1989, he had held his mother
captive in her home in Battle Creek for a week, beating her, police
said.
His mother refused to testify against him, and he accepted a plea
agreement.
In 1996, she suffered a broken pelvis when she fell from a third-
story window in Battle Creek. She said it was an accident.
I'm afraid you're mistaken about the sample being a "simple," "pure" blood
sample. This was clearly stated to be a mixed sample of at least two
donors, with all the added complexity that fact entails. I am not a DNA
analyst, so I am not qualified to argue the statistics with you (I'll leave
that to those so qualified). However, I will point out a seeming flaw in
your logic if not your math (again, I leave the latter to others). You seem
to assume that because a profile is not proven unique that there will
necessarily be false matches. That is a non sequitur conclusion, and
although I admit I have not been following this thread closely (and so have
not digested all of your presentations) I have not seen how your
"mathematics shows that there _must_ be false matches occurring." Can you
explain for the non-statisticians among us exactly how your math "proves"
that false matches "MUST" occur? I know enough to realize that no
statistical calculations with any number of loci can ABSOLUTELY eliminate
ALL possibility of a coincidental match with more than one person (there is
still that "1" in whatever astronomical number), but that by no means
indicates that such a coincidental match MUST occur. With sufficiently
large statistical probabilities, that possibility becomes so vanishingly
remote as to be practically nil.
Further, when you not only consider the statistical probability of one in
billions or trillions (or more), but also consider the question of access to
the crime scene (i.e., only a limited segment of the population could have
access to a given crime scene, not the entire world population), it
magnifies the statistics even more, making the probability of a coincidental
match in the _relevant_ population even more remote. For example, if a rare
profile would statistically be expected in, say, 6 people in the world
population, but 5 of those six people happen to live on the other side of
the world and have never visited the country the crime occurred in, then
within the relevant (local) population, statistically that profile would be
expected to be found in only one person, not six. As other list members
have pointed out in past discussions, the relevant question is not how many
others could share the same profile but rather how many others WITH the same
profile could have deposited the sample at the scene. DNA statistics should
be considered within the context of all the other evidence and facts of the
case, not in isolation. Besides, DNA can only prove a connection between a
person and the recovered evidence; it doesn't prove the person committed the
crime (there may be an innocent explanation for the connection).
Finally, while I am again not qualified to argue the numbers with you, I
could not help but notice that the calculations you presented in the
messages I did read concentrated on 9 loci, yet you try to extrapolate the
weaknesses you see in those statistics to 13 loci, as if adding four more
points of comparison increased the numbers by only 4/9 or 44%. But each
additional loci increases the statistical numbers by much, much more than
that. I know enough about statistics to understand that using 13 points of
comparison instead of 9 increases the statistical probabilities by several
orders of magnitude.
The crux of the problem is that forensic scientists,
on continually hearing figures like 640,000,000,000
or 844 septillion etc, have got it in their heads that
DNA profiles show uniqueness.
Once you have, even CODIS databases, in the low millions
the mathematics
shows there must be false matches appearing.
Its just you cannot predict, prescisely, who will
be implicated falsely. Just that those with all
high allele frequency profiles will be likely 'framed' first.
It does not require billions .... septillions.
It does not require contamination, sloppy or
inept sample handling/processing.
I thought it was a pure crime-scene sample of blood
not a mixed sample - again fundamental.
There is nothing wrong with DNA evidence that isn't wrong with every other kind
of evidence. Heisenburg showed that there is no such thing as 100% certainity.
In fact, we have to rely on circumstantial evidence, until we invent a time
machine videorecorder and can go back and tape the crime. This argument is
purely academic.
Dear Gerrit,
Thanks from all of us down under in OZ who also are
aware that we are far from perfect....I have always
had a problem dealing with the 'super EGO's' that only
the USA can belt out with such gusto...!!!! I have
never been so happy to be an Aussie in all my life and
can't quite believe that these two supposedly
professional people would behave like this in a public
forum....
Cheers from Oz
Are there cases where DNA alone is used to convict ? I
know of many where it has been used to exclude.
I don't recall any cases where DNA has been used without some surrounding
circumstancial evidence to support it. But a defendent's blood on the body of a
murdered person without other identifying evidence, could be enough to convict,
I suppose. Really hard to explain its presence to a jury.
"In three of them only Leiterman's DNA was indicated.
The other two showed it was a mix of DNA identified as
Leiterman's and Mixer's, he said. "
I found no reference
to what form the 'Ruelas' trace was found. Implicated
via database so implication is uncontaminated single
trace - AFAIK mixed stains can never lead to
a single large database match as usually thousands of possibles.
Has there ever been a 46 peak , on 13 loci, two person
mixed stain result ? I very much doubt it.
Derivation of formula for minimum number before
repeats start occuring.
Start with a loaded dice analogy,
then convert to the directed pairs situation and finally
convert for the more varied locus/allele situation.
Consider a 10 faced loaded dice with weighting
such that
face 0 or face 1 have a probability of 0.2 each
face 2 or 3 , probability 0.15 each
and faces 4 to 9 , 0.05 each
Toss 10 times and record the 10 digit number
Repeat n times.
Determine a number N where a repeat
of a previously occuring 10 digit number will occur.
The probability of a random pair of single
digits matching is
Sum of the squares= 2(.2^2) + 2(.15^2) + 6(.05)^2 = 0.14.
The digits
in each of the 10 positions are independent, so the overall
probability of all 10 digits matching is (sum of squares)^10 ~= 2.893e-9, and call p.
To generate N numbers, there are N(N-1)/2 pairs of numbers which
must all be different to avoid a repeat. If the pairs were
independent then the expected number of repeats would be pN(N-1)/2,
which will be 1 when N is about 26,000. This estimate for the
expected value should be fairly close for N << 1/p.
so N = SQRT(2/p)
By comparison, if
the numbers were unbiased , normal dice, then about 1 repeat in the
first 140,000 numbers.
It is impossible to predetermine what matches/repeats will
occur , just the number of tosses before they start to
appear. Doing this on computer gave
"0101026021"
"1012303012"
"3153031131"
as 3 pairs of repeats in a 30,000 run. Do another run
and they will be different figures and second point is the
matching numbers do not exclusively only contain the
most common weightings of 0,1,2 and 3 , there is the odd
5 and 6.
Now convert to factor-in directed pairs, the dice are tossed in pairs
and the results written down smaller,larger, analogous to 2 alleles
per locus.
The factor p now becomes (2 * 0.14^2 - 0.14^3)^5.
The -0.14^3 term because 'homzygous' pairs are invariant
on directing.
eg "3153031131" becomes directed to "1335031113"
likes of "1335301131" also becomes "1335031113"
A 30,000 run gave 21 , 10 digit directed-pair matches
Now convert to the DNA profile situation and formula becomes
For n loci 1..... 5 (6,9,10,13,15 or any number)
and m (valid) alleles at each locus and 2 per locus.
So Allele Frequencies are AF1 ..... AFm
Let Sn be the sum of the squares of AFs at locus n
ie Sn = AF1^2 + AF2^2 +...... + AFm^2 for each n
Let Qn = Sn^2 for each n
Let p = (Q1 * Q2 * .... * Qn ) [(2-S1) * (2-S2) * .... * (2-Sn)]
Then N = minimum number before finding a match is
N = SQRT (2/p)
As in the DNA profile situation you cannot predict
what the matching digit numbers will be , just that they will
consist of the more common alleles. Someone with a
number of rare alleles is very unlikely to ever have
an unrelated match. With emphasis on unrelated. If
fathers "play away" and there are unknown half-siblings
with some rare alllele also appearing in such a database then
that changes matters.
Summary for minimum number of profiles for the totally
artificial situation of 'parthenogenic' profiles with no 'parents'
all totally randomly generated
6 loci pre-2000 FSS, UK Caucasian: 9,900
10 loci FSS, UK Caucasian: 2.3 million
9 loci NIFS, Australian Caucasian: 420,000
9 loci NIFS, Australian Aborigine: 290,000
13 loci CODIS, USA Caucasian: 20.5 million
The above and below in agreement with simulation
results.
Calculated minimum numbers for sub-sets of profiles like
mine of > 8 per cent AFs , also for >3 per cent ,Codis case.
6 loci UK Caucasian for >8% then 2,400
10 loci UK Caucasian for >8% then 280,000
9 loci Australian Caucasian for >8% then 44,000
9 loci Australian Aborigine for >8% then 21,400
13 loci USA Caucasian for >8% then 2 million
13 loci USA Caucasian for >3% then 10.2 million
All my alleles have an AF greater than 8.6 %, it
is that subset ie all >8% that are in most danger.
I have on order, from the British Library, a publication
that may allow me to put a minimum AF distribution
to a population to discover what proportion of a
population substructure/co-ancestry would have minimum AFs
of 8%, proportion 5%, 3% etc
Everyone's DNA profile is rare until it has a match
in multi-million databases. In my definition of "rare"
an individual would have a number of rare alleles, say
sub 1 percent AF. Then his chance of false implication
is much much less than someone like myself. I don't
have a Lithuanian or Iroquois grandparent to take me
out of the background high AF genepool and this
Damoclean situation. Now false matches are appearing
in these large databases it will occur with more and more
frequency. There are now millions of profiles
in these databases and thousands of match-check searches
a week. The cross-over point has now been passed where
there is no guarantee that a consequential match is
real rather than false.
Re my last post on derivation of false match formula
my 13 times table is rusty and should have referred to
4 * 13 = 52 peaks and not 46
The blood evidence incriminating John Ruelas was
stored away 36 years ago. That historic evidence and
his present day personal samples have been separatly tested
and repeated again.
There is no doubt that the DNA profile matches -
the problerm is forencsic 'scienmtists' who have a
mindset that believes false matches are impossible
because they keep hearing, and repeating in court, nonsensical
numbers like 43 trillion to one ... even up to 840 septillion
chances against such events.
Equally damning there is no validation research
into the non-transmuteability of samples stored
at ambient, or frozen for that matter, for 30 years or more.
When OB/GYN physician office labs were found to be
using there grandmothers to view pap smear slides in
the 80's, the Clinical Laboratory Improvment Act was
passed in '88. Unless similar standards in education
and certification are enforced , human error will
always be the greatest factor in mistaken test
results.
Keep in mind that the vWA results as YOU see them are YOUR opinion
and no one elses on this planet. I have stated to you at least once
that YOU CANNOT MAKE ANY JUDGEMENT REGARDING THE CMJ article PLOTS.
Since you put forth a fundemental tenet of science, let me give you
another. Sceintist don't make statements (such as yours) without
interpreting all the data. Yet you continue to assert your opinion
of the CMJ data without all the facts. Not to mention the fact you
obviously have no training or experience which would qualify you to
do so. You have clearly read just enough information on the subject
to make you dangerous.
With respect to reproduciblity... the science is reproducible and all
the validations and papers written on the validations demonstrate as
such. So your statement is baseless. Stutter does not mean a
properly trained analyst can't interpret data. You are wrong...
period! Leave this science to those who are trained to perform it.
Regards,
Ken
P.S. For the love of God can we please end this thread. The horse
is dead.
Dear Ken,
It is best to ignore this guy and continue your belief in what you know. It
is so very tempting to blast him when he types with such ridiculous
statements, but that is taking the bait and letting him run with it. The
delete" button is good way to deal with him.
Are you saying electrophoegrams are inadmissible
in court ?. You would like, no doubt, the situation where just sets of
numbers are presented in court after being fiddled, sorry
interpreted.
I repeat the science is not reproduceable.
from FSI 143 (2004)
5 profiles in 2055 individuals, not mixed stains, were
falsely homozygous on vWA
3 in 2055 false H on D8
2 in 2055 false H on FGA
3 in 2055 false H on D18
2 in 2055 false H on D5
so 0.7 per cent wrong due to false homozygosity
I am not aware of the validation studies into falsely
included stutter. Please inform me where you've found
that research to show 0 per cent error rate on stutter.
I agree there are some very dangerous people around
this subject. I have recently been contacted by someone
related to a prisoner
denied parole in the UK because his DNA profile matched
a cold-case DNA profile from 15 years ago.
The proveable fact he was abroad at the time of this
historic crime is ignored because the UK's dangerous
characters have a mindset that places DNA 'evidence'
above normal exculpatory evidence. Having been contaminated
by repeated hearing, if not saying, false match figures
like 840 septillion to one. No court case, just continued prison
sentence, because the UK's dangerous characters have
deemed him guilty with no other evidence whatsoever.
Mirroring this case in USA
http://www.sltrib.com/utah/ci_2491514
19 Dec 2004
<...>
Rudy Michael Romero was about to be paroled for an armed robbery conviction when he was linked to a series of rapes along the Jordan River Parkway in the early 1990s.
Based on recent DNA tests, the Utah Board of Pardons and Parole revoked Romero's July parole date and ordered he serve at least another 25 years.
While DNA cases that free prisoners have become common, the Romero case is raising eyebrows in the legal community - though nobody is rushing to his defense.
Utah's parole board acknowledged it has entered "uncharted waters" by keeping Romero behind bars based on evidence of crimes for which he has not been convicted. Romero will never be tried for the Jordan River rapes because the four-year statute of limitations has expired.
But the DNA evidence was compelling enough for parole board members to deem Romero a sexual predator who would pose an "unacceptable risk" to the community.
Parole Board Chairman Michael Sibbett says that because Romero is serving a five-years-to-life term for aggravated robbery, the board can keep him locked up for the rest of his life.
End Quote
As I am not a DNA analyst I am not knowledgeable about the effects and
distribution of the phenomenon known as "stutter," so I can not and will not
attempt to debate that with you (I again leave that to others, if there are
any with any patience left). However, the very fact that you seem to expect
a total absence of variance in a scientific measurement clearly demonstrates
how very shallowly you understand the science you so passionately declaim
(and prevaricate about). I am somewhat amazed that you can be impressive in
your mathematical musings, yet make such ignorant statements about the
larger picture. This suggests an inherent (and insurmountable?) prejudicial
bias in your viewpoint that is blinding you to some very basic scientific
maxims.
Let me try to explain where you've gone astray this time. Science is not as
straightforward as simple mathematics. There is variance in ANY measurement
of any kind, because no measurement can be perfectly exact. A measurement
can be accurate but not precise, precise but not accurate, both accurate and
precise, or neither. Accuracy is determined by comparison with known
standards, and precision is calculated based on the variance observed
between successive measurements. Therefore the concept of reproducibility
is relative, not exact. Scientific measurements are not stated with
absolute certainty but rather within a set of given "confidence limits," or
"degrees of uncertainty," e.g., a weight might be expressed as "125.045g
+-0.006g." I believe this is what others were trying to tell you when they
spoke of having to interpret the raw data to determine if and where stutter
occurred, and how it might affect (or not affect) the results. Simply
because a predictably variable phenomenon like stutter occurs randomly in a
measurement process is no indication that the method is unreliable or
"unscientific." Variation is normal and expected in any measurement
process. Such phenomena are studied and characterized, so that they can be
accounted for in the interpretation of experimental results. This is why
expert human interpretation is an essential ingredient in any analysis.
If you find it difficult to grasp or accept the inevitability of measurement
variance and/or the difference between accuracy and precision, try an
experiment. Put away your calculator for the nonce and pick up a yardstick
(or meter stick); then use it to measure the length of a football field and
see if you ever get exactly the same figure twice, down to the inch or
centimeter (you won't, because there will inevitably be some variance in
exactly where you place the stick each time in the many placements needed to
measure the length of the field). Does that mean the measuring stick is
unreliable? No, it does not. The yardstick may be exactly 36 inches (or
the meter stick exactly 100 cm) in length and it certainly doesn't change
during use, therefore it is accurate; and if your use of it is skillful, so
will your measurements be accurate. Yet you will naturally and unavoidably
have some variance in your application of the stick to the field and so will
get varying totals each time you sum your group of measurements. This is
measurement variance or imprecision, and it is inevitable. Do enough
repetitions of these groups of measurements, and you will be able to
determine the mean, degree of variance and standard deviation of your total
set of measurements (you can pull out your calculator to do this), and so
will be able to accurately state the length of the field within those given
confidence limits (i.e., the mean will be accurate within the stated degree
of precision). For example, you may come up with a figure of something like
99.53 yards, plus or minus 0.9 yards. That is a statement of both the
accuracy and precision of your conclusion.
This is how scientific measurements are determined and described.
And with that, a good weekend to all, and to all a good night!
I'm glad to see you are checking primary sources.
But it pales into the background, for errors,
compared to this more general study
that should also be consulted
Int. J. Legal Med. 2004, 118:83-89
The Gednap blind trial results from 2002 over
30 European countries. It gives a good idea of the
sum of errors including transcription, misinterpretation etc.
Forgetting for the moment the forensically artificial
use of 50/50 blood stains, also blind rather than
double blind ( they knew they were being tested ).
On the other hand there were, deliberately, a
couple of hard inclusions
FGA (42.2) in with 21,22 and 23
and THO1: 8.3,9,9.3,10 which often needed
human interpretation.
I've not been able to get the GEDNAP protocols
and the published table on p84 does not state the
number or average number of samples sent to
each lab. The earlier 1990s tests, also in the table,
were certainly tests per locus and from the text there
was at least 5 stains sent out.
Year 2002 results for 136 different European labs and selected
from a pool of 17 loci there were 30,479 (single locus rather than
multi-locus ?) tests, giving an error rate of 0.4%.
My interpretation is - one or more wrong alleles in each of
these single locus tests recorded as one failure. But as 50/50 contributor
tests it is also a reasonable approximation to expected single
contributor profiles errors.
If anyone is aware of a similar interlab test for single
samples I would be pleased to get the reference. From
the text it was not possible to determine if they were all
mixed samples or the labs knew this in advance.
So scaling to multilocus situation and FSS 10 loci
would be something of order 4% of all profiles erroneous in one allele
and 5% of all profiles for CODIS.
From an earlier discussion,
and consensus, on this group there has to be 20 matching FSS
numbers or 26 in CODIS before declaring a match between
crime-scene and database. Only fair because choosing ANY
9 from 10 plus one of the remaining pair increases the false
match probabilies by very large factors, let alone for CODIS.
As I am not a DNA analyst I am not knowledgeable about the effects
and
distribution of the phenomenon known as "stutter," so I can not and
will not
attempt to debate that with you (I again leave that to others, if
there are
I keep being told that stutter is recognisable if
given the raw data, whatever that is compared to
an electrophoreogram unless they just mean skewness
in the printing process can displace the traces
from the ladder on the original straight off the
machine plots.
No one is saying what the signature or
characteristic of a stutter trace actually
is, compared to a real allele trace.
A very simple question I would have thought - is it
breadth of trace, sharpness of peak, off-ladder
response always tied to a specific on-ladder allele ?.
Put simply what are trainee analysts told to
look for ?
So for those professing to be knowledgable here,
what is wrong with the following definition ? It
seems fair enough to me - but what do I know.
http://forensicdnaconsultant.com/Glossary.htm
Stutter- an artifact of STR amplification, in which the n-1 allele (n is the true allele; n-1 is the true allele minus one repeat unit) produces a signal that is 5-10% of the peak height of the true allele. For some loci, this can be as high as 20%. Stutter peaks can rarely occur at the n-2 and very rarely at the n-3 position.
Returning to
http://www.cmj.hr/2003/4403/18DROBNIC.pdf
p351, Fig 1, Plot B, 2 in from left of bottom line
the unlabelled anomalous vWA peak would appear
to obey the n-1 or n-2 or n-3 rule as
on the lower weight side. But to my reading it is way above
20% of 15,16 or 17 peak heights.
Also without extraneous
input what marks out this vWA(*) peak as being an artiface,
rather than a real allele?
Still everyone keeping quiet on this black art.
I agree with Adam. It is also important to remember
that Paul feels that these little exchanges are part
of some "peer-review" (his words - see the bottom of
http://www.nutteing2.freeservers.com/dnapr.htm) to
validate his agenda. You're never going to have an
open discussion where Paul sees the other point of
view.
Please quote correctly
"'peer reviewed'" not "peer reviewed" or is the
difference too subtle. Or is it again two nations
divided by a common language ?
Here is the response from the paper's author...
"Dear mr.Howard,
if you look carefully the electropherogram of AmpFlSTR SGM Plus, you
will see no extra peak at vWA locus in Figure 1 plot A, this extra
peak you mention (Fig.1 plot B) is result of pull-up from the other
color in yellow because the concentration of DNA used in PCR was to
high (RFU is between 4000-6000, but recommendation is 1500 to 2000).
best regards,
katja Drobnic"
So as Michael Coble stated, it is pull-up from D3.
So Mr. Nutteing, in all fairness you should add a correction to your
website's mention of this article. It should state that you are dead
wrong, there are no "extra" peaks that can't be interpreted, and
there is no malfeasance in the reported case. I would just delete
all mention of the article from your site a few days after the
retraction is posted.
I hope this matter is now completely resolved.
Well I, at least, have learnt something. It would be an
excellent test-piece for a trainee analyst, if it was not
already on the internet.
So a rider should be added to this definition of stutter
http://forensicdnaconsultant.com/Glossary.htm
Stutter- an artifact of STR amplification, in which the n-1 allele (n
is the true allele; n-1 is the true allele minus one repeat unit)
produces a signal that is 5-10% of the peak height of the true
allele. For some loci, this can be as high as 20%. Stutter peaks can
rarely occur at the n-2 and very rarely at the n-3 position.
Rider:
The reference to 10 to 20% factors only relate to a
reference sample situation. With exagerated amplification,
to bring up minor contributor peaks, then any stutter could be
exagerated disproportionately, way above 20%.
Or is the consensus that this CMJ vWA(*) was not stutter at all
but bleed-through ?
Returning to the CMJ case adapted for the more normal situation
of a mixed stain sample, with strong major/minor differentiation
and just a victim reference sample. So just plot 1B and Figure 3A
in effect (ignoring amelogenin ).
Then as far as I can see the only indication in that case of
vWa(*) stutter/anomaly would
be specifically the coincidence of this vWA(*) with D3 (18) and D5 ( 12)
which I certainly didn't spot and a general caveat concerning
over amplification. IMHO an over-worked analyst in that
circumstance may have declared that the suspect would have
this spurious vWA(*) allele.
How many analysts, routinely, have the benefit of cross-referencing
between Power-plex and SGM+ determinations ?
DNA matches & exclusions/Where do you draw the line?
While Paul only focuses on the problems with false accusations (due to poor
analysis - not fault of the premise), should someone not point out all the
cases that are reopened for DNA analysis. The defense (defence?!) requests
DNA analysis to overturn the conviction done before DNA was so frequently
used. Many people have been released based on this new DNA evidence. By
Paul's theory, many of these were released without true cause. If DNA
matches are not consistently correct, why would DNA exclusions be correct?
Should we go back to the age before DNA analysis and let falsely accused
people get convicted due to circumstancial evidence or ABO typing (or
similar technology)? If the statistical analysis is wrong in one case
(DNA) would it not be wrong in another (ABO blood typing)? Blood typing
uses statistics to determine matches and exclusions - although not on such a
grand scale. Both technologies should be treated similarly. So if we
cannot use statistical analysis in our search for guilty parties, what are
we left with? Eyewitnesses? The proverbial smoking gun? Confessions?
Statistical analysis said that my fiance would survive about 5 years after
treatment - the fact that he lasted only 8 months from diagnosis does not
mean the statistics were wrong. It merely means that statistics cannot
predict the future. There are always 'outliers' I was told by my statistics
professor. There are exceptions to almost any 'rule.' My fiance's genetic
predisposition for melanoma was statistically negligible, but he contracted
it nonetheless. Are those statistics wrong as well? Where do you draw the
line, Paul?
IMHO - this is all poppycock. Statistical analysis may not be perfect, nor
may DNA analysis, nor may DNA analysts, nor any branch of forensic science;
they are better than incomplete analysis.
I hope for the sake of society that you do not work in the criminal field. I
find your attitude insufferable and inexcusable. However, you are not the
exception, sorry to say.
No, we do not live in a perfect world. However, if DNA is found to be
objectively imperfect then it should be treated no differently than a hair found
at
a crime scene. It should be given no more credence than any other
"circumstantial evidence". It should not be allowed to pass the Kelley-Frye
test.
As conservative as I might be I do believe it is far better to let ten (10)
guilty people walk free than to wrongly incarcerate one (1) innocent person
regardless of the nature of the crimes involved. That is one very good reason
why the United States of America is, by far, superior to all of the other
nations of this world. Let's try to keep it that way.
Greetings from Belgium, whose inhabitants know it's imperfect.
I find *your* attitude insufferable and inexcusable. You seem to be jumping
to conclusions without any facts - just an instinctual dislike of me
apparently. I'm merely making a point/asking a question, yet you pounce on
me without a second thought (or dare I say even reading the entire email).
The point being - if people are being falsely accused, people are also being
falsely exonerated.
Has DNA been found to be objectively imperfect? To my knowledge all 50
states here in the US, most (if not all) European countries, places in
Africa, basically all around the world DNA is accepted as a high quality
means of identifying perpetrators.
It is my belief that people should not be falsely accused nor falsely
exonerated. If this is happening (which Paul has not proved by any means)
then DNA analysis does need to be called into question. However, until this
is the case, Paul's theories are just that - theories. Very few inmates
will acknowledge that they commited the crime in question - however, in most
cases the system worked properly in incarcerating them - When pressed
closely, many of the inmates will tell one another or their case-worker that
they were correctly convicted.
DNA analysis has been extensively used by defense attorneys to exonerate
their clients after they've been convicted. It is unfortunate that innocent
men/women are convicted, but as I said previously - this is not a perfect
world. However, if DNA analysis is being used to exonerate falsely
convicted people - DNA analysis was probably not used to convict them.
Therefore, DNA analysis is doing just what you want it to - it is ensuring
that those who are convicted are truly responsible for the crimes they are
accused of.
I am a forensic biologist - a good one (just as my supervisor)! I do my job
and will go above and beyond the call of duty. Something of which you
have no knowledge. Could it be a guilty conscience speaking for you? My
analysis of evidence is very conservative and correct. I will never say
something is so that has not be thoroughly tested - and even then, I make my
statements very conservative (not "Human blood of type blah blah was found
on Exhibit blah blah which completely matches Suspect blah blah).
However, I believe that the scientific community as a whole and the forensic
science community in specific has thoroughly investigated all the tests that
are currently used in medicolegal situations. I do not question my
phenolphthalein results because it has been tested and retested and accepted
by the scientific community and forensic community (defense attorneys
included). DNA is still fairly new, but still a sound science. Innovations
will take place and make it more accurate, easier to accomplish, cheaper to
perform, etc.
By the way, previously you used similar language to object to my spelling
correction of Paul (defense and defence) (you apparently didn't read that
entire email either because no mention was made of Paul's error in calling
dyslexic's "sloppy minds"). You indicated that I was being closed-minded
and superior because I was only acknowledging the American spelling.
However, at the end of this most recent tirade you say you think American's
are superior because "the United States of America is, by far, superior to
all of the other nations of this world." This certainly doesn't sound like
you are being open-minded. Maybe you should be consistant in your opinions
and practice what you preach. Maybe you should avoid personal attacks
altogether and stick to the facts and failing that - just avoid reading my
emails if they bother you so.
This is an open forum for discussion of forensic-science. I am discussing
forensic-science and asking questions of Paul to try to understand his
viewpoint. Before you attack - you may want to thoroughly know your subject
and what is being said.
I mprisoned by DNA - freed by ABO blood groupings ?
So it goes on - Australia this time. You couldn't invent this stuff
and be believed.
http://www.heraldsun.news.com.au/common/story_page/0,5478,12051494%
255E1702,00.html
Quote
Murder pardon being sought
By Nikki Todd
25jan05
A MAN convicted of the murder of a British waitress on a remote Great
Barrier Reef island more than 20 years ago could find out within
weeks whether his petition for a pardon has been granted.
Sydney businessman Wayne Edward Butler was jailed for life in
Queensland in February 2001 after DNA evidence linked him to the
murder of Celia Natasha Douty, 41, on Brampton Island in the
Whitsundays in 1983.
Ms Douty's bashed body was found in September 1983, in scrub land
behind a beach where she had been sunbaking.
A red towel draped over her body bore blood and semen samples which a
jury almost two decades later agreed was linked through DNA evidence
to Butler.
But Butler, who was holidaying on the island with his wife at the
time of the murder, maintained his innocence throughout his trial and
subsequent appeal.
Now 61, Butler has lodged a plea for pardon with Queensland Governor
Quentin Bryce.
The pardon is being considered by the state's Crown Solicitor, with a
recommendation expected to be delivered to the Governor within four
weeks.
The pardon is considered Butler's last resort, with none of the 55
other petitions lodged by prisoners in the past decade in Queensland
proving successful.
Butler's pardon application is based upon new evidence by forensic
scientist and blood group specialist Professor Barry Boettcher, who
questions the validity of results found by the Queensland
government's forensic lab, the John Tonge Centre.
Prof Boettcher, who has been involved in high-profile cases including
the exoneration of Lindy Chamberlain and the Peter Falconio murder
case, believes Butler should never have been convicted, claiming his
blood group did not match the semen sample.
"It is my opinion that, since the ABO blood grouping of the semen
stains excluded Wayne Butler as being the donor of the semen, there
had to be an error in the DNA results," Prof Boettcher wrote in his
final report, according to this week's Bulletin magazine.
The Bulletin article says Prof Boettcher is convinced, after
examining information gained via Freedom of Information, that
Butler's test tubes were mislabelled at the John Tonge Centre.
Butler's sister Alison Butler said her brother did not hold much hope
he would be released.
"I don't think Wayne dares to hope," Ms Butler told the Bulletin.
"He had faith in the system. He wanted to go through the system
because he thought he would be exonerated.
"He was suffered a wrongful conviction because of a bungle in the
laboratory and it has wrecked his life."
End Quote
Chimeras
Blaschko lines are chevron type alternating patterns that appear in
skin pigmentation associated with chimera
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=pubmed&dopt=Abstract&list_uids=2667350
Am J Hum Genet. 1989 Aug;45(2):193-205.
Quote
Association of pigmentary anomalies with chromosomal and genetic mosaicism and chimerism.
Thomas IT, Frias JL, Cantu ES, Lafer CZ, Flannery DB, Graham JG Jr.
Department of Pediatrics, University of Nebraska Medical Center, Omaha.
We have evaluated eight patients with pigmentary anomalies reminiscent of incontinentia pigmenti or hypomelanosis of Ito. All demonstrated abnormal lymphocyte karyotypes with chromosomal mosaicism in lymphocytes and/or skin fibroblasts. In seven the skin was darkly pigmented, and in all of these seven cases the abnormal pigmentation followed Blaschko lines. The literature contains at least 36 similar examples of an association between pigmentary anomalies and chromosomal mosaicism, as well as five examples of an association with chimerism. The pigmentary anomalies are pleomorphic, and the chromosomal anomalies involve autosomes and sex chromosomes. The pigmentation patterns are reminiscent of the archetypal paradigm seen in allophenic mice and demonstrate the clonal origin of melanoblasts from neural crest precursors. Patients with anomalous skin pigmentation, particularly when it follows a pattern of Blaschko lines, should be appropriately evaluated for a possible association with chromosomal or genetic mosaicism or chimerism.
End Quote
DNA testing question
My copy of
FSI article 112 (2000) p151-161
is in paper form only, after
application to the British Library
technical journal requests.
In UK terms 1.50 GBP plus 0.10 GBP per page
I don't know what the LoC charges would be
but I do know that a lot of USA academics
use the BL, Boston Spa , Yirkshire UK
facility in prefernce to the USA system
IIRC for reasoins of turn around speed
rathe rthan costs.
AFAIK all countries have a similar system
for otherwise expensive subscription access.
Certainly US, UK and Oz do
I returned to FSI 112,2000, 151-161
Happened to notice in the sub-study for cross-validation
of different somatic samples from same individual.
"Any stutters that occured in the profiles were below
10 per cent of the associated allele and were not prevalent
in any sample type."
I would suggest this troublesome anomalous vWA(*) peak
in amplitude or area terms is supstantially more than that
declared 10% threshhold.
Translating to a hypothetical single contributor sample
then if homozygous (h,h) in whichever allele
causing the stutter, I would posit the 'interpretor'
would have declared a result vWA (*,h) not vWA (h,h)
Also in that FSI 112 study the nearest I've seen to long term
validation
research. One 17 years old blood sample stored at ambient temp
and RH. High molecular wt peaks especially D18 and D2
were degraded almost to mush-level compared to the remainder
also attenuated but useable. No mention of whether this adult was
traced and typed for migration/anomalies due to
UV damage or whatever. I find it absolutely abhorent
that people are being incarcerated on 20 or 30 yearold DNA
'evidence' with no validation studies into this as far as I
can find ,qv (Guthrie cards thread ).
Returning to FSI 143 (2004) 47-52
showed 5 samples in 2055 having false homozygosity
So once again in routine, this time single contribution samples
how does the person doing the 'interpretation' handle that.
Results by these (French) analysts/interpretors
were 2 homozygous in 15 and one each at 16 , 17 and 18
As far as I can see only discovered as false homozygosity
precisely because it was a validation study using 2 different systems.
True ( or at least different on othe rsystem) vWA results
were (15,17), (15,18) ,(16,18),(17,18), (18,21) all
falsely homozygous on the low allele.
I seem to have made an error on the D21 result , the others are correct,
for Bobby the
chimpanzee and DNA profile using 'human' SGM+ kit;
Amel. X,Y ; D8 (10,12) ; D21 (23,24) ; vWA (12,12) ; FGA (17.3,23.3) ; D3 (14,14) ; THO1 (6,7) ; D2 (21,22) ; D16 (9.3,9.3) [ 8 +1 loci ]
From same table 1 on page 156 is an unnamed female gorilla
DNA profile with null alleles on D21
Amel. X,X; D8 (7,7) ; vWA (17.3,17.3) ; FGA (27.3,32.1) ; D3 (16,17) ; THO1 (4,7) ; D2 (22,24) ; D16 (11.3,12.3) [ 7 + 1 loci ]
[ Remember just 6 loci was considered perfectly adequate by UK FSS
'scientists' to be able to quote uniqueness figures of 1 in 37 million
in courts, to prosecute, during the 1990s. ]
Other primates had many null alleles.
Of other more distantly related species the 'best' performing was a
domestic cat giving two responses
one peak on D8 12.1 and one peak on D3 19.3
Justice institute seeking answers on the reliability of fingerprints
Study urged on quality of evidence, comparisons
By Flynn McRoberts and Steve Mills
Chicago Tribune
Originally published February 21, 2005
Four years after scuttling a study into the reliability of
fingerprinting, the research arm of the Justice Department is seeking
answers to fundamental questions about the granddaddy of forensic
science.
The National Institute of Justice recently called for researchers to
explore such crucial issues as how to measure the quality of
fingerprints lifted from crime scenes and the accuracy of comparisons
made by law-enforcement examiners.
The research solicitation seeks to "provide juries with increased
information about the significance and weight of fingerprint
evidence" and also to create tools "to improve the fingerprint
examination process," said Catherine Sanders, spokeswoman for the
Office of Justice Programs, which includes the institute.
The agency's decision is the latest example of an unmistakable shift
in the previously defiant world of fingerprint experts. Until
recently, they had pointed to nearly a century of convictions in U.S.
courts to dismiss calls for a closer examination of their discipline.
The institute's solicitation is "very significant because it's
recognizing that there is an area that needs to have research done,
and they're willing to step up to the plate and fund it," said Ronald
Singer, president of the American Academy of Forensic Sciences. "What
I think NIJ is hoping is that they can fund the research that will
validate the process and answer a lot of the questions."
Across the nation, law-enforcement and forensic officials are
beginning to acknowledge the need to show whether science supports
the expert witnesses whose opinions have long been accepted as gospel
in American courts.
The broader reassessment of fingerprint comparison is largely being
driven by a series of high-profile errors committed by examiners,
including their role in the wrongful conviction of Stephan Cowans, a
Boston man imprisoned for six years after a bogus match linked him to
the shooting of a police sergeant.
A few months after Cowans' release last year, an even more
embarrassing mistake occurred when the fingerprint world's elite --
examiners at the FBI lab -- falsely connected Brandon Mayfield, an
Oregon lawyer, to the Madrid train bombings through a print found
near the scene.
For several years, the National Institute of Justice has sought to
walk a fine line over fingerprint comparisons -- seeking to bolster
the scientific foundation of a time-honored discipline, without
admitting weaknesses.
Calling for any research gives ammunition to defense attorneys who
can suggest that the very need for such a study shows that the
discipline is not entirely reliable.
Robert Epstein, a federal defender in Philadelphia, did just that
several years ago when his client was charged in a robbery. In one of
the only cases where a judge has told prosecutors to prove the
scientific validity of fingerprint comparison, the FBI asked crime
labs across the country to help the bureau convince the judge, in
part by examining fingerprint evidence from the case.
While most examiners agreed with the FBI's conclusion that the
defendant's prints matched those found on the getaway car, 17
examiners in nine states were unable to make an identification --
underscoring that the discipline is much more subjective than many
fingerprint experts have acknowledged.
After receiving the conflicting responses, one of the FBI's top
fingerprint experts asked the dissenting examiners to take another
look, with the help of some FBI enlargements of the prints in
question.
"These enlargements are contained within a clear plastic sleeve that
is marked with red dots depicting specific fingerpint
characteristics," wrote Stephen Meagher, chief of the FBI lab's
latent print unit, in a June 1999 letter. "Please test your prior
conclusions against these enlarged photographs with the marked
characteristics."
(Contacted last week, the FBI declined to elaborate on the letters,
saying they were part of the public record.)
Three months after Meagher's letter, the National Institute of
Justice approved a call for research into fingerprinting, only to
eventually let it die amid an uproar from police and prosecutors.
But continuing questions about the reliability of fingerprint
comparison and the experts who practice it increasingly are prompting
crime labs to re-evaluate how they do their work.
The FBI has long left it up to individual examiners to determine
whether they have enough points of comparison -- among the loops,
whorls and arches that make a fingerprint -- to identify a match.
But the bureau is now considering imposing guidelines on them,
including the hundreds of examiners who evaluate tens of thousands of
prints a day at the FBI's fingerprint database headquarters in West
Virginia.
"Probably within the next year, we're going to set our own standards -
- a minimum number of points needed to declare a match," said Charles
Jones Jr., a fingerprint examiner instructor at the FBI's West
Virginia facility.
Asked why the bureau was considering the change, Jones said: "I guess
the Mayfield case was an eye-opener for everybody."
Noting that the examiners at the FBI lab outside Washington, D.C.,
have "always been the cream of the crop," he added, "If those guys
can make a mistake, so can we."
End Quote
false positive ten locus dna matches, more or less
I'm doing some research and trying to find false positives such as
Peter Hamkin or the Goettingen case where there have been false
positive dna matches at ten or so str loci. I am especially looking
for news reports, journals, or other published sites besides
websites
by afficionados.
All help would be most appreciated.
Someone with my own concerns - very much a rarity.
I don't know how many loci, as that sort of info does
not seem to leak into the general media.
The 'Watters' one was 6 loci.
All of the
below cases were considered matches sufficient to
arrest and seriously inconvenience to say the least -
except the ones with the perfect alibis of course.
If the news-agencies have deleted any of the following
links then I have archived them myself, but you will
have to email me for that.
Notice the followup investigations , if any, are never released
into the public realm - I wonder why.
I am the only person who has reported this
material in its entirety. Le Monde published
a piece that could well have substantially been derrived
from my material but that is all in the established
media public or likes of FSI or IJLM.
Anyone know of the followup to the John Raelas
4 year-old 'rapist / murderer' ?
Obscure because supressed.
The FBI expunges false matches from databases
before alllowing others to look
"Budowle testified (2) that
all the matches have been edited out of these databases and that this
removal is justified because it is not possible for two individuals
to yield identical profiles when as many as seven probes are used."
Science,Vol 256,26 June 1992 p1743
Author Patrick J. Sullivan
A Nick Davies crime/feature reporter of the UK national
newspaper The Guardian was going to write a piece but
something/someone (D-notice ? ) warned him off.
In the space of one week Special Branch closed down
4 of my web-sites to suppress. I had to contract back
to my Russian site. They had to arrest me on a b.ll.cks
charge in a completely failed attempt to keep me quite.
3 of those 4 sites regained and something like 9 more
started in direct response to this corrupt form of
attempted censorship.
Please feel free to update but I doubt there is any
change in principle - non disclosure. The UK FSS
are just the same in 2004 / 2005 - total non-disclosure.
More complete quote from the original letter
Fom Science,Vol 256,26 June 1992 p1743
Author Patrick J. Sullivan
Title : DNA Fingerprint Matches
I am writing to comment on two aspects of the report " On the
probability of matching DNA fingerprints " by Neil J. Risch and B.
Devlin (7 Feb,p717) . Risch and Devlin searched several large
databases to determine whether there were any samples with matching
patterns across a nummber of gene loci. They found " the probability
of a matching DNA profile between unrelated individuals to be
vanishingly small....."
Last summer I was trying a Federal Bureau of Investigation (FBI)
case, Minnesota v. Johnson (1),and examined three FBI databases,C-3
(Caucasian),B-4 (black), and H-3 (Hispanic). During my examination,I
discovered 25 apparent matches. Before my examination ,the existence
of these matches had been known by only a few individuals connected
with the FBI. Bruce Budowle of the FBI subsequently testified in
Minnesota v. Johnson that he was aware of these matches and that they
had been discovered when the FBI examined its database with its
computer matching program. The FBI was able to verify that most of
these matches occured because the Texas College of Osteopathic
Medicine submitted more than one blood sample from the same
individual. One false match was the result of sample handling error.
The FBI also discovered three sets of matching samples from Florida.
These samples were from the black and Hispanic databases. The FBI was
not able to identify that the Florida matches were the result of
duplicate submissions from the same individual or of submissions from
identical twins. Budowle then asked Cellmark Diagnostics (German-
town,Maryland ) to examine the matching samples. Its probes also
yielded unclear results. The Florida matches were then deleted from
the databases,even though there was no explanation for their
occurance.
The FBI again revised its databases in January 1992. The new
databases are designated C-4,B-5, and H-4. Budowle testified (2) that
all the matches have been edited out of these databases and that this
removal is justified because it is not possible for two individuals
to yield identical profiles when as many as seven probes are used. My
first point is this: Of what scientific value is a paper that seeks
to draw any conclusion from the fact there are no matches in a
database when the matches have been removed from the database before
the analysis is done? The FBI's removal of matches from its databases
before giving them to outside scientists guarantees that those
scientists' conclusions will support the FBI's "self-fulfilling
prophecy."
This is not an isolated practice. Budowle testified in United States
v. Yee (3) that the FBI ran its match program over its South Carolina
black database and found a large number of matches. The FBI's record-
keeping was such that it could only speculate as to the cause of
these matches. Again,the FBI removed them from its database.
End Quote
These are interesting points. The problem remains that in countries
like Canada, for instance, there is no uniform policy mandating
forensic labs or prosecuting attorneys to use 13-locus matches only in
court. As a case recently showed in Chicago, this can lead to false
allegations or arrests.
I am dealing with a case where there is a nine-locus match and the
prosecutor is refusing to assist in having the evidence sample tested
at 13 loci. This case also concerns the possibility of fraternal
contributors. I'm hoping to raise the awareness level of the
possibility of nine or ten locus matches, whether between brothers,
relatives, or non relatives.
If anyone can assist me, I'd be most gracious. I have followed many of
the links in Paul Nuetting's website. As a result, however, I am still
looking for documented nine or ten locus matches that were cleared up
by more discriminating tests. I am also looking for information
documented in news, papers, official websites and the like as to what
the standards are in European or other country forensic databases
regarding the number of loci used for their records. I know that
England is 10. I've heard that Italy is 13. Some reliable sources are
needed.
Thanks very much for all responses,
You will find its universally accepted within forensic
science Oz 9loci, UK 10 loci or USA 13 loci that you
do not get false matches. They have so blinded themselves
quoting nonsense unfounded false match probalities such
as 640 billion .... 840 septillion. They always say it must
be cross-contamination or some such impondrable.
I am in contact with a relative of a UK prisoner denied
parole because his DNA profile matches a crime-scene
profile obtained 15 years ago despite him being abroad
at that time.
You would think it was one of the most
fundamental of researches to be published
in FSI or whatever.
This was the last disclosure on false matches
back in 6loci days.
Forensic Science International 95 (1998) p30.
Concerning data in the UK DNA database as of 04 October 1996
when there were only 6311 samples from the London
area and 573 from the Cardiff area.
"A small number of unresolved duplicate pairs of
profiles were present in the regional data :10 pairs
within the London region and 1 pair in Cardiff.
The most common cause of
duplicate entries is the use of aliases by suspects
who have been arrested on several occasions.
For administrative reasons ,it is not always possible
to resolve such duplicates by exhaustive
police investigation."
It could not be simpler to cross-correlate with
mug-shots or dermal fingerprint database to
check for aliases. I wonder why they didn't
or at least publish the results.
Please contact any relevant persons in the FBI , FSS, NIFS
etc to run their match checker over the arrestee database
and disclose the numbers now the totals are in millions
They only get reported in the mass media, never
in forensic science journals. I had it from a leading
forensic scientist that this Prof Chaseling 7 loci false
match in 5,500 and other such matches
was to be published in likes of FSI but never was
(in archive as newspaper site is now subscription)
http://web.archive.org/web/20001025074133/http://www.smh.com.au/news/0004/22/text/review8.html
Don't want to bite the hand that feeds it. Its
all about clear up rates for minimum expense.
No matter prosecutions via false matches, it
is 'justice' on the cheap which is the main aim.
I wonder how many FF database custodians are actually privvy
to the number of unrelated (non- alias)
10 loci matches in that section
of the FSS NDNAD. That is the fundamental figure
that even parliamentary questions will not elucidate.
Odd to think that general press is more open
than the forensic press such as FSI,IJLM, JFS,JFSs.
The one exception I can think of is
International Journal of Legal Medicine (2004 ) 118: p83-89
German DNA Profiling
Group , blind trials of testing samples at 136 labs in 31
European countries. Despite using unrealistic 50/50
mixed samples the error-rate expressed in terms of
10loci DNA profiles - 4 percent are erroneous.
How many forensic scientists are aware of that one?
No analysis of multi-million arrestee-section databases
for false matches which of course is the best way of
establishing the true number of false matches between
crime-scene profiles and arrestee database.
No mention of the database numbers after which
false matches are bound to appear just using allele
frequency information.
6 loci UK Caucasian approximately 5,000
10 loci UK Caucasian approx 300,000
9 loci Australian Caucasian approx 50,000
9 loci Australian Aborigine approx 25,000
13 loci USA Caucasian approx 2 million
I previously quoted the published number of 6 loci false
match numbers (10 in 6,311 + 1 in 573) . If none of those were due to aliases then
these above minima may be reduced even further by a factor
of 10 or so as it reflects co-ancestry .
In other words we have to believe the authorities are quite
happy to have multiple criminal records for criminals ,
so they can choose which previous to be used against them.
Acres of forensic science journal pages happily multiplying
numbers together to produce off the planet numbers
conveniently ignoring that it only applies if inheritance
is independent across different loci, ie no cross-linking
of loci/alleles.
I was recently reading this weighty tome
History and Geography of Human Genes - Cavalli-Sforza et al
The Sardinian population comes top in the world in allele frequency terms
for inheriting each of these genes
MSN*M , HP1 , HLAA*18, HLAA*2 , PGD*A , DIA 2
Remember the newspaper reports concern a match between
a crime scene and an arrestee-side database record.
Matches contained within such databases or otherwise
hidden until there is a repeat of the R v Watters (6 loci)
situation of a crime-scene profile matching an existing
unrelated database pair , so 3 way match.
The newspapers do not have to report the 'science'.
They just report the results as it affects people.
Peter Hamkin in 2003 was arrested and forcibly removed to the other
side of the country for extradition hearing on charge
of murder. He was implicated by DNA database match that was
sufficient in UK and Italian administrations to show
'guilt' to be arrested and transported.
The Goettingen prisoner would have been arrested
because of a false match in the BKA
database under the German administration, but
inconveniently he was in prison at the time of the murder.
John Ruelus in USA would have been charged with rape
and murder, unfortunately he was 4 years old at the
time.
These are the results of 'scientists' blindly
going along as the church did pre Nicholas Copernicus
and Galileo Galilei.
Well a false match is one possibility that does not need
twisting the facts to fit the case or blaming
incompetence/errors in testing. No mention of this possiblity
of course.
Part Quote
http://www.timesdispatch.com/servlet/Satellite?
pagename=RTD/MGArticle/RTD_BasicArticle&c=MGArticle&cid=1031781820698
Doubt raised about implicated man
BY FRANK GREEN
TIMES-DISPATCH STAFF WRITER Mar 28, 2005
Last year, DNA testing cleared Arthur Lee Whitfield of two 1981 rapes
in Norfolk and implicated Aaron Doxie III, an imprisoned sex offender.
Now things are not quite so clear.
Authorities have turned up evidence raising doubt about the
culpability of Doxie, who is serving three life terms for unrelated
sexual assaults in 1984.
The unusual turn of events pits the sometimes questionable
reliability of eyewitnesses against the work of the state crime lab,
which is under scrutiny for its work in the case of Earl Washington
Jr., who was nearly executed for a rape and murder he did not commit.
Not only are the two rape victims in the Whitfield case sticking to
their identification of Whitfield, but both said their assailant was
not circumcised. The two were assaulted in such a way that they would
presumably know.
It turns out Doxie is circumcised; Whitfield is not.
If the women are not mistaken, it would raise questions about the DNA
testing or evidence handling -- or both -- performed by the Virginia
Division of Forensic Science.
Whitfield was freed on parole in August after the DNA tests. But
problems DNA arose for him after he took the formal step of
petitioning the Virginia Supreme Court to prove his innocence.
<...>
Witch-hunt targets scientists (Australia)
http://www.thecouriermail.news.com.au/common/story_page/0,5936,1264316
7%255E3102,00.html
Witch-hunt targets scientists
Hedley Thomas
24mar05
SCIENTISTS at the John Tonge Centre are being threatened with jail in
the wake of a government hunt for the source of leaks highlighting
serious problems in the forensic laboratories.
Queensland Health has ordered a sweeping review of the centre after
leaks to The Courier-Mail this month forced an official admission of
problems and sparked a review of the centre.
Sources close to the centre said scientific staff were "furious and
close to walking out in disgust" after being told to get independent
legal advice and prepare for questioning.
Senior bureaucrats will begin a probe into the leaks as early as
today.
Scientists have been warned a police investigation is probable and
any unauthorised disclosure of information about the troubles at the
centre could result in prosecution and imprisonment for two years.
News of the inquiry, which was condoned by Health Minister Gordon
Nuttall, was leaked yesterday within minutes of scientists being
given formal written notices warning of the ramifications.
"They are upset and feel intimidated," said a source at John
Tonge. "They are getting advice on whether they will even turn up.
The department wants to include past and present staff in the
inquiry."
Scientists met yesterday and took advice from the Queensland Public
Sector Union before deciding to flatly refuse to participate in any
investigation.
Mr Nuttall's office confirmed the probe, describing it as
an "internal audit around the information provided to The Courier-
Mail".
Mr Nuttall, who has responded to revelations about equipment and
accuracy problems and a huge backlog by ordering an external review
and starting the outsourcing of scientific testing, yesterday refused
to comment on the crackdown.
In a statement, Queensland Health said: "It is usual practice for
Queensland Health's internal Audit Branch to investigate any
unauthorised release of confidential documents as such actions are
breaches of the Queensland Health Code of Conduct."
Staff, who have been heartened recently that public disclosure of the
problems is finally leading to some ministerial action, are furious.
"It has a (management) culture that refuses to accept there's a
problem - a culture where everyone who dares to criticise is written
off as an idiot," said Ron Grice, a former senior scientist at the
centre.
"Staff are demoralised, traumatised and angry. The backlogs there
today are worse than three years ago when I left despite the millions
of dollars put into the place. Morale is at an all-time low. I often
wonder why the whole place doesn't just implode."
Opposition Leader Lawrence Springborg accused the Beattie Government
of embarking on a reign of terror "to silence public servants and
whistle blowers with legal threats and intimidation".
"Rather than fix the problems and answer the questions, they go out
to get even with the people who showed them up," he said. "That's
clearly the consequence of raising any concerns about this
government - they will get even with you and fit you up."
Mr Beattie, who leaked Crown Law advice about himself last week, told
Parliament yesterday that he valued "open, accountable and
transparent processes", but it was wrong for public servants to leak
material.
The leaks inquiry will try to determine who provided The Courier-Mail
with an internal document exposing flaws in the testing of human DNA.
DNA QUESTION
I am not a DNA analyst, nor do I care to be one. LOL Therefore, I
am
not sure this question will make sense, but here goes: When
dealing
with a STR or a VNTR are the STR or VNTR on a specific chromosome
or
are they "fragments" from a number of chromosomes?
Thanks for your help,
These are the 10 STRs used by the UK Forensic
Science Service (apparently)
vWA
THO1
D6S502
FGA
D21S11
D18S51
D2S1338
D16S539
D19S433
D3S1358
plus the Amelogenin X /Y notifier
I forget which chromosomes that vWA,THO1
and FGA are variously on but for the remainder the number
after the D is the chromosome number.
Moot point about what you mean by specific.
Notice D6S502 in the above list , that is what they
put on their official forms. But unbeknown to
themselves, it would seem, what they actually
analyse and record is D8S1179 on chromosome 8
and not the D6 one.
Would you Adam & Eve it ? Supposedly forensically
admissible documentation.
DNA Databasing
Yes, it was routine for long term storage of biological samples to
be in a freezer in 1984. That was the year I began training as a
crime scene investigator, and the two long term storage options that
I recall were drying out the sample or freezing it. I don't know
that aren't validation studies of DNA stored for 20 years, they
would be simple enough if the contributors of some of the stored DNA
were still alive.
There is not even validation study into biological
material that has been frozen for 20 years and then
tested against current samples from the same contributors.
Let alone study of stored material at ambient conditions.
In FSI 112 (2000) page 157 one study into one 17 yearold
blood stain stored at ambient
which showed the 'heavy' loci D18 and D2 responses so
depleted over 17 years as to be almost forensically unuseable.
Responses of about 3000 rfu for D19 gradually
reducing in peak heights down to D18 and D2
peaks of about 150 rfu . Even that single reported analysis
did not have a comparison with a current sample from the
same person to check for any changes.
By pure coincidence I saw this report today
concerning loss of frozen state for just 4 days.
Quote
http://www.abqtrib.com/albq/nw_local/article/0,2564,ALBQ_19858_3705941
,00.html
Freezer error prompts query from victims
More than 60 cases are connected to the temporary meltdown in the
Police Department's evidence room.
By Joline Gutierrez Krueger
Tribune Reporter
April 16, 2005
Officials with the District Attorney's Office are attempting to allay
crime victims' fears that their cases could be in jeopardy because of
a freezer malfunction in the Albuquerque Police Department evidence
room.
But whether the victims' cases are among the more than 60 cases
District Attorney Kari Brandenburg says are involved in the temporary
meltdown two months ago, advocates say the fallout could be
emotionally wrenching for victims.
"It's tragic because it not only affects those victims whose cases
might directly be connected to the evidence that is destroyed but
every victim that has a pending case out there," said Elena Giacci,
executive director of the Coalition to Stop Violence Against Native
Women. "They start worrying: `Could it be my case? Could my attacker
go free?' It brings up the whole trauma again, the emotion that's
attached."
Brandenburg says her office has begun to review a list of 1,644 items
police said were in the evidence room's faulty freezer to determine
what cases were affected.
A list of between 60 and 64 cases was completed Thursday, she said.
Of those, most are still pending or have recently been tried or
pleaded out, she said.
At least 15 are murder cases, and nine are rape cases, she said.
Sandy Dietz, director of the District Attorney's Office Victim Impact
Program, said she and her victim advocates are contacting every
victim or family member whose case is on the list.
"People are worried until we begin to explain to them exactly what
happened and exactly what is being done," Dietz said.
Prosecutors involved in each case are also speaking with victims and
families, she said.
The freezer went down during the weekend of Feb. 19. The problem was
discovered Feb. 22, and the freezer repaired Feb. 23, Brandenburg
said she learned from a police memo.
A freon leak caused the shutdown, said police Capt. Larry Sonntag,
supervisor of the Metropolitan Forensic Science Center.
<...>
End Quote
All the following is just speculation for consideration.
The DNA sections chosen for profiling are
in the so-called "junk DNA" areas, where not only are there
DNA repeats, but on human evolution time-scale there
has been enough mutations to be of use for identity
purposes. To me this would suggest that such DNA
is not as robust as the "active" DNA. Errors in this
DNA , due to stress conditions for example, would not be significant
for biological inheritance. But errors in the junk DNA, due to
freezing and un-freezing,
will not necessarily lead to unviable dogs or humans
but could mess up adduced profiles.
Another consideration, harking back to a fundamental
assumption. To multiply allele frequencies together, an
implicit assumption is made, that all loci/allele inheritance
are independent of all others.
The "junk DNA" may
be involved with (computer parlance)
error correction/parity checks cross-
coupled to "active" DNA.
http://www.nature.com/news/2002/020916/full/020916-4.html
by Donall Mac Donaill , subsciption now, or discussed on
http://www.idthink.net/biot/error/
Freezing DNA nonarevers
I've checked back to early 90s of indexes/abstracts
of FSI & IJLM but nothing found for
validation along these lines. Next weekend I will
go further back if they are on line.
On freezing
All the following is just speculation for consideration.
The DNA sections chosen for profiling are
in the so-called "junk DNA" areas, where not only are there
DNA repeats, but on human evolution time-scale there
has been enough mutations to be of use for identity
purposes. To me this would suggest that such DNA
is not as robust as the "active" DNA. Errors in this
DNA , due to stress conditions for example, would not be significant
for biological inheritance. But errors in the junk DNA, due to
freezing and un-freezing,
will not necessarily lead to unviable dogs or humans
but could mess up adduced profiles.
My understanding of the law (probably UK and USA)
is evidence of this nature can only be submitted in
court if validation checks of a scientific procedure
have been thouroughly
researched and preferably published in peer-review.
People have been prosecuted where the sole evidence
of any import is DNA profile analysed from cold-case
samples - some stored at ambient and some frozen.
Also people exonerated using DNA profiles from
cold-case samples , equally suspect.
There would seem to be no justification for this, ambient
or frozen, and any cases
processed in court are then open to be appealed.
Bellow is a link to a recent presentation by my
colleague, Margaret Kline, about her experiences with
long-term storage of DNA. Several references can be
found within the presentation. Margaret has been
testing several stains at various conditions
(including liquid nitrogen -150 oC... now that's
freezing!). Stains stored at room temp will begin to
show some degradation over time (no shock there), but
typeable DNA was recovered from stains at all samples
examined.
As far as potential errors in junk DNA regions due to
freeze/thawing, here is my speculation...
Let's suppose that a tube containing 1ng of DNA
undergoes several rounds of freeze/thawing. Suppose
that the genotype at TH01 in the fresh sample of this
individual is 8, 9.3. In 1 ng of DNA, there are
approximately 167 total pairs of chromosomes (one cell
= 6pg of DNA; 1 ng = 1000pg / 6pg = 166.7 cells).
Thus, to believe that freezing and thawing can affect
the genotype, on the 167 chromosomes that have an 8
allele at TH01, nearly each and every locus would have
to spontaneously have to "mutate" (due to less overall
robustness - not sure what that means) to say a 7
allele. Meanwhile... the other 167 chromosomes will
"mutate" from a 9.3 allele to... I don't know 9...
10... 11... you get the point. In a sea of 3 billion
nucleotide bases only a thimbleful of bases changes
occur in the junk regions (I'm not sure of the
mechanism here since the cellular enzymes have been
inactivated or removed during extraction) of
forensically chosen markers to "mess up" genetic
profiles.
I think it's easier to believe that DNA can degrade
from repeated freeze/thawing by shearing rather than
the spontaneous generation of novel alleles.
I was interested to hear what people think of forensic science and
how
accurate it is. I have read there have been a few mistakes in the
lab
but should we rely on as much as we do.
I am reminded of the quote, attributed to John Wanamaker
"I know half the money I spend on advertising is wasted, but I can
never find out which half."
Whenever areas of forensic science are subjected to
the rigours of blind tests and validation things start
coming unstuck. False match statistics for DNA profiles,
quoted in courts, are often plain weird unsubstantiated numbers
like 640 billion to 1. Subject DNA profiling to multi-lab
validation then it emerges that something like only 96
per cent of 10 loci DNA profiles are correct ( even worse
for 13 CODIS structure ).
See (abstract)
http://www.springerlink.com/app/home/contribution.asp?
wasp=b82ee8788c4f4b9da0b00953db41ac04&referrer=parent&backto=issue,4,1
3;journal,9,53;linkingpublicationresults,1:101167,1
In normal use no one knows whether a particular DNA profile
is in the 96% or the 4% group.
In court presentations of traditional fingerprints even
a layman juror, given two non-smudged or skewed enlarged fingerprints,
can compare for himself the pattern of termini, bifurcations and enclosures.
Not so with DNA profiles - an abstraction from variations at the
molecular level. You cannot give that same juror a microscope
so he can see for himself the repeat patterns of C,G,A and Ts.
All you can say is that these DNA profiles are self-consistent
in that repeating the processing produces the same results.
You cannot state with universal truth that that is the real underlying profile/s.
Is that homozygosity on vWA, D8, FGA,D18 or D5 real or
an artifice of the system you are using ?
FSI 143 (2004) 47-52
False Homozygosities comparing 2 DNA profiling kits with samples taken from
2055 individuals showing 15 errors between SGM plus, Powerplex 16
and Profiler systems.
And then of course the more general operational errors
highlighted by multi-lab validation
International Journal of Legal Medicine (2004 ) 118: p83-89
0.4% of 30,479 single locus validation tests were wrong
translating to something like 4 to 5 percent of UK or US profiles.
No way is that truth.
False matches involving unrelated people will not occur
in the first instances with people with rare alleles.
The first to occur will be people having all 18/ 20 / 26 alleles
being greater than 5 per cent allele frequencies.
Of all the random simulations of large databases no
matches have occured where any of the (AF) allele frequencies
are less than 5 per cent. Each of my 20 are greater than 8 per cent
compared to ethnic background.
When I start to see court presentations where the minimum AF data is
expressed, or similar analytic criterion, then I will know that they are not
spouting unsubstantiated Alice in Wonderland twaddle such as
1 in 3 quintillion
1 in 695 quadrillion
1 in 17 quadrillion
1 in 1.82 quadrillion
1 in 25 trillion
I've yet to see a court presentation where consideration is
made of whether a defendent could be one of these first
tranche of all high AF profiles that can easily be falsey matched in
low millions of profiles , long before billions , trillions etc.
Another USA false DNA profile match - updated
Independent confirmation but I have an easy explanation of course.
http://www.freep.com/news/metro/leiterman21e_20050621.htm
Part Quote
UNRAVELING A MYSTERY: Old memories, old killing, new twists
June 21, 2005
<...>
How a drop of blood found on Mixer's hand that belonged to a then-4-
year-old boy will factor into the case is still unclear. The blood
belongs to John Ruelas, 40, of Jackson, who was convicted in 2002 of
beating his mother to death.
DNA evidence from the Ruelas murder was being processed in the
Michigan State Police DNA laboratory in Lansing on the same day in
2002 that evidence from the 1969 Mixer case was also being processed.
Police have been unable to connect Ruelas to Mixer or Leiterman. The
fact that DNA was being processed on the same day raises questions
about possible lab contamination.
Results of an independent test of the blood, requested by Hiller,
shows it belongs to Ruelas, Gabry said. An evidentiary hearing is
scheduled for Monday to determine whether an independent analysis of
DNA from both Leiterman and Ruelas, requested by Gabry, can be used
at trial.
Leiterman remains in the Washtenaw County Jail awaiting a July 11
trial.
no mention of using SGM instead of cofiler or whatever,
I wonder why
DNA expert agrees with bizarre results
Hearing in 1969 murder case adds to confusion surrounding evidence
Tuesday, June 28, 2005
BY ART AISNER
News Staff Reporter
The testimony during a pretrial hearing did little to clarify the
confusing evidence presented so far in the case
against Gary Leiterman, charged in the 1969 death of Jane Mixer.
Dan Krane, an associate professor of biological science at Wright
State University in Ohio, testified at the
request of Leiterman's defense attorney. Krane said he agreed with
the state police lab's findings that DNA
from blood samples on Mixer's body belonged to Mixer, Leiterman and a
convicted murderer named John
Ruelas, who would have been 4 years old at the time of Mixer's death.
Leiterman, 62, of Gobles, is charged with murdering Mixer, a 23-year-
old law student from Muskegon who
was found shot and strangled in a western Wayne County cemetery in
1969. He was arrested last November
after police matched his DNA obtained after a prescription fraud
conviction with DNA found in stains on
Mixer's pantyhose.
His trial is slated to begin July 11.
The pretrial hearing is scheduled to resume next Wednesday with
testimony from another expert in DNA
analysis and an official from the crime lab, both called by the
prosecution.
The DNA test results came under scrutiny from Washtenaw County
Circuit Judge Donald Shelton after a
May hearing when Deputy Chief Assistant County Prosecutor Steven
Hiller confirmed Ruelas' DNA was
present in the Mixer sample and that evidence in Leiterman's case was
tested on the same day as evidence
from a murder investigation involving Ruelas, 40, of Jackson.
There was no testimony Monday regarding the possibility of
contaminated samples at the State Police
laboratory. Hiller declined comment after Monday's hearing but has
said he does not believe the evidence
was contaminated.
Defense Attorney Gary Gabry on Tuesday did not say he could prove
otherwise, but noted the initial test
results from the lab did not indicate a mixture of DNA samples or the
presence of DNA from a minor. He said
he only discovered that information upon Krane's insistence to review
all the documentation related to the
tests, including the technician's notes and underlying data.
"I'm troubled by the concerted effort to not make all the information
available,'' he said. "If they believe in
their opinion, why not provide all the information that's there?''
End Quote
Ho-hum here we go again
Why do I still keep seeing such reports?
http://www.macleans.ca/topstories/news/shownews.jsp?content=n061443A
June 14, 2005 - 17:11
Defence expert slams RCMP handling of DNA in 30-year-old murder case
AMY CARMICHAEL
VANCOUVER (CP) - A blood sample used to link murder suspect Robert
Bonisteel to the stabbing deaths of two 14-year-old Vancouver girls
in 1975 was mishandled by the RCMP, a DNA expert said Tuesday.
The speck of blood found on Bonisteel's shoe, smaller than a dust
particle, was kept in unsealed bags that could easily be
contaminated, said Dr. Ron Riley. "It's so easy to contaminate a
sample with such a small amount of DNA," said Riley, who was
testifying for the defence in B.C. Supreme Court.
"It happens in many labs, it's written about all the time. It's a
limit of the technology. The sample in this case is just so small I
don't think it's scientifically significant. To use it as the
foundation of a probable match, I think that's a mistake."
Bonisteel is charged with first-degree murder in the stabbing deaths
of Judy Maria Dick and her friend Elizabeth Inge Zeschner.
It took 26 years of testing the sample and considerable advances in
DNA science to finally obtain a partial profile suggesting the blood
on Bonisteel's shoe belonged to one of the teen victims.
The sample, just six or seven human cells, degraded over time and was
too small to be reliable, said Riley.
"They're going below the threshold used by other labs and that's
risky. It's such a small amount I don't think it's reliable."
According to lab reports, RCMP scientists placed the fragile matter
in uncapped tubes in a spinning centrifuge, right beside samples from
the victims. This is done to concentrate the samples, but they should
be sealed during the process, Riley said.
Court has also heard the samples were handled by staff who didn't
change gloves between touching the victims' bra and sweater and then
picking up Bonisteel's shoe.
The Crown accused Riley of having idiosyncratic views not shared by
anyone else in the scientific community. Experts for the prosecution
have told court the odds of the analysis being wrong are one in 14
billion.
Even with all the alleged mistakes, Riley said the likelihood that
the blood on Bonisteel's shoe could belong to someone other than Dick
is one in 3.3 billion.
Bonisteel left Vancouver shortly after the bodies of Dick and
Zeschner were found in suburban Richmond in February 1975.
His marriage had just broken up and he abandoned his wife and five-
week old son. His lawyer James Millar now says Bonisteel was "in a
great deal of turmoil" at the time.
On the way to Winnipeg, he raped a woman. Once he reached the city,
he raped again. He spent 20 years in jail for the attacks.
Prosecutors allege he killed the 14-year-olds who were so close they
said they were sisters.
Police launched a massive undercover operation to try to get a
confession.
An investigator posed as the boss of a biker gang in which Bonisteel
was desperate to climb the ranks.
The undercover cop told him the police had DNA evidence on him
linking him to the murders of Dick and Zeschner. He said Bonisteel
had to tell him what happened if he wanted the bike gang to help him
make the problem to go away.
Bonisteel's lawyer said his client was afraid of the undercover cop
who he believed was a ruthless gang leader.
Millar said Bonisteel just told him what he wanted to hear.
The undercover officer videotaped a long statement by Bonisteel, who
described holding a knife in one girl's chest while the other sat
terrified in the front seat of his car. He then sliced the other
along her rib cage, "because it was the nastiest thing" he could
think of.
End Quote
Interesting story. I had to read the centrifuge portion twice before
I figured out what they were talking about. It sounds as if they
were concentrating samples in a speed-vac or with microcon-100's. In
the case of the speed vac then yes the samples would all be open at
the same time because you can't close them. However, I agree that I
would not concentrate suspect and victim samples together. The
chances of contamination during this procedure are incredibly remote
and I know of no instances of it. Despite this I maintain that I
wouldn't do it. It just looks bad even if it really isn't.
As to the experts statement that only " six or seven cells" were in
the sample, it sounds a bit unrealistic. That would be approximately
36 to 42 pico grams of DNA. I doubt with extended PCR cycle times
and going below threshold you could even obtain a "reliable" partial
profile given the age of the sample and the fact the sample has had
previous testing, which would decrease the DNA available for current
testing. Our labs minimum guide of 200 pico grams, using normal
analysis parameters, generally gives a partial profile at best or a
weak complete profile.
As to the "contamination" issue. What does sealed mean? A piece of
tape on a paper envelope is considered sealed. If that sample spent
the entire 20 something years in an unsealed package then they had
better have a really good excuse for it or pay the price by a defense
reaming. And the glove issue is just simply embarrassing. I never
open suspects items while victim items are open. I always change
gloves even moving to items from the same individual or even during
examination of a single item. We lay our evidence items out on clean
butcher paper with new paper for each item. My entire work surface
is wiped down with either bleach or RNase between evidence items from
different subjects. If I even think for a second I have contaminated
my work surface then I will wipe it again prior to opening a new
evidence item. Let's just say that I burn through gloves, especially
during extraction/clean-up.
I hope that my comments illustrate the point that not all labs are
incompetent and we actually do care about what we do. I don't want
to wrongly implicate or exclude someone because of sloppy work.
676 quintillion
In the UK, this week, a professor Roy Meadow is being pilloried for
criminal abuse of statistics presented in court. Do any
forensic statisticians live in the real world ?
Multiplying AFs together is just as ridiculous as
what Prof Meadow is accused of doing.
http://www.delawareonline.com/apps/pbcs.dll/article?
AID=/20050623/NEWS01/506230316/1006
By CHARLOTTE HALE / The News Journal
06/23/2005
<...>
But what O'Neill didn't know before a preliminary court hearing
Wednesday was how steep the odds are that the DNA belongs to someone
else.
Newark police detective Andrew Rubin testified during a preliminary
hearing that 18 zeros are needed to create the ratio: one in 676
quintillion.
<...>
Self-consistency ?
Who but forensic scientists could make the following reported statements
and find a consistent whole ?
"DNA testing of the dried blood last year found it matched another convicted murder, John Ruelas. Ruelas, however, was just 4 years old at the time of the murder and lived in downtown Detroit, a 45-minute drive from the crime scene."
"The lab that handled Mixer's evidence also handled Leiterman's DNA and that of the boy, John D. Ruelas, now 40, who is serving time for the beating death of his mother."
"He also said there will be highly technical testimony on DNA and that experts will tell jurors that the odds of the samples belonging to a Caucasian male other than Leiterman are less than one in 170 trillion. "
"Prosecutors acknowledged that samples in both Leiterman's and Ruelas' cases were tested at the lab on the same day but have denied claims of evidence contamination. "
How dare they present such a dog's breakfast to the court
and counter-GIGO fashion expect the court to sort it out.
170 trillion indeed.
So the John Ruelas connection is due to unrelated false match.
http://www.mlive.com/news/aanews/index.ssf?/base/news-
13/1121870476251160.xml&coll=2
Wednesday, July 20, 2005
...
Leiterman's attorney, Gary Gabry, has maintained that Leiterman never
knew Mixer. and that the two were linked only through questionable
DNA evidence. What has yet to be explained is the DNA profile of
another man, 40-year-old John Ruelas, was in a spot of blood that was
found on Mixer along with Leiterman's DNA profile. Ruelas was 4 years
old when Mixer died. He is now in prison for killing his mother in
Jackson County.
Sylvia Gill, a DNA analyst with the Virginia-based Bode Technology
Group, a private forensic lab, testified that her company was asked
by the Michigan State Police crime lab to conduct DNA testing. The
results, she said, showed Leiterman's DNA samples matched a partial
DNA profile on Mixer's stocking.
Megan Shaffer, an assistant director at ReliageneTechnologiesnear New
Orleans, another private company that does DNA testing, testified
that the test results of a DNA extract of blood off Mixer's left hand
were consistent with those of the MSP crime lab. Shaffer said the DNA
tests excluded it from being Mixer's own blood.
Previous testimony in the case showed that MSP lab tests found
Ruelas' DNA on Mixer's left hand.
From "millions of trillons more times likely" to zero likelihood
No other scientific discipline is allowed to make such nonsense
statements without full account of errors.
When are forensic 'scientists' going to start telling the truth
in courts of law ? . The true probability is far more like 1 in 20
that any 13 loci DNA profile determination is erroneous. That is
the only publically available error rate, from all possible sources of
error, for forensic labs blind tested (GEDNAP blind trial concept part 2
,data for year 2002, pub 2004. No single lab from this validation
study or any other inter-lab exercise has published their own results, so this
pan-Europe reusult is the only one to go by.
Until that error rate is up
in the trillions to 1 against then such statements are totally bogus and fallacious.
John Ruelas last week and now the latest pile of nonsense reported here
http://www.reflector.com/local/content/news/stories/2005/07/26/20050726GDRlincoln.html
...
"inherently lacking the scientific foundation and safeguards required in constitutionally protected criminal trials in North Carolina, especially in the absence of strict assurances designed to stop the SBI lab from repeating past mistakes and reporting erroneous results with little integrity and no consequences to the SBI lab or agents responsible."
...
"It's pretty shoddy stuff that's coming out of the lab," Savage said. "They're not doing what anybody would do in what I would call a real lab. If you're going to put someone on death row, put someone in jail for life, or even for a year, at least the science needs to be sound."
I would put it a lot stronger than that.
While laboratory error rates are of significance, no one to date has
a reliable way of measuring them. Utilizing one study to base lab
error on is both foolish and irresponsible. The statisical
significance of an STR profile is, in my opnion, not directly related
to lab error rate. I believe it is a go/no-go concept. If there is
an error then the STR frequency is irrelevant until the error is
identified and understood in the context of a case. The best we can
do is develop procedures that allow for the detection of error.
Also, the procedures should minimize the chance of error.
So a true probability of 1 in 20 is bogus and fallacious. Do you
have some math supporting this? I have read several papers on lab
error rate and proposed caluclations to modify STR freqeuncies. All
the calculations are based on theoretical lab error rates. However,
I still maintain that error rate and STR frequencies are mutually
exclusive. I believe lab error rates needlessly attempt reduce the
statistical significance of a STR profile. Assuming for a moment
that one could accurately measure a labs error rate then the
following would be fair: The frequency of occurance for a given
profile is "A"; however, the chances the lab obtained that result in
errror is "B". Good luck with determining "B" though! Not to
mention... what is an error? Please define all categories of error
that would be used to generate this calculation.
You hammer away at the statistics we use, yet you are willing to take
one study and apply it's results in a broad stroke. This seems
inconsistent to me. Shouldn't something be well characterized before
application?
Delighted to explain.
If you are aware of any other inter-lab validation
test conducted in blind fashion please tell me.
I don't mean internal validation checks as in these terms
that is meaningless as using same reagents/machines etc
The relevant table of results from the only proper
published study I am aware I've placed (fair-use fashion ) on
http://www.nutteing2.50megs.com/gednap.gif
( if remote linking to an image does not work it is
on a link about 1/2 way down on the URL in sig piece)
It shows 0.4 per cent error rate in 30,479 single locus
tests.
Combined results from 136 labs in 30 European countries and multi-locus
tests across 18 autosomal STRs and 12 Y-STRs.
published in
International Journal of Legal Medicine (2004 ) 118: p83-89
http://www.springerlink.com/app/home/contribution.asp?wasp=df08a8b230994088a1e8983c28228a00&referrer=parent&backto=issue,4,13;journal,10,54;linkingpublicationresults,1:101167,1
or
http://tinyurl.com/cnquy
The GEDNAP blind trial concept part 2.
This data is anonymised so the good and bad are lumped
together - considering the bad do not know they are
bad (or at best, in disagreement with the consensus).
Only one error per profile on one locus (1 or both alleles)
is sufficient to invalidate a profile.
Simple maths 0.4 *13 = 5.2 per cent or 1 in 19 , say 1 in 20
for Codis , 1 in 25 for NDNAD structure
Until more such proper inter-lab validation studies
are undertaken and published this is the only series
and the only data to work with. We can quibble about
1 locus per profile or 5 per profile or whatever but this
report didn't divulge. I can come back by mentioning
that the stains were, forensically almost unheard of,
50/50 blood/blood stains. It is still an extremely long way
short of 1 in trillions error rate.
If practising forensic scientists have not obtained and
read this report and others on false homozygosiies etc
then shame on them.
All labs say they are error free , because they just
don't know otherwise until such blind exercises are undertaken.
Then surprise - surprise the bad ones in this GEDNAP
process were informed but has anyone seen them mention
their individual results - NO.
No links to other relevant inter-lab validation studies heard.
So until there is then this Gednap is the best available,
don't blame me, I can only use what is available.
There was qualitive analysis of the errors discovered but
no quantitive breakdown of the "0.4%" figure.
Error types
Sample transposition
Transcription errors
Previous studies showed marked errors from certain
types of stain so surprise,surprise they settled on
the least error prone of fresh blood source only.
Wrong interpretation of minor allele and stutter.
FGA (42.2) often unrecognised
Distinguishing THO 8.3,9,9.3 and 10
They had deliberately used the rare FGA (42.2) but there
was no mention of deliberately including homozygous
pairs known to produce false homozygous results.
Separate study on this showed FSI 143 (2004) 47-52
Same sources processed on SGM plus and Powerplex
and compared.
15 errors in 2055 single contributor profiles not otherwise
selected to have the following error-prone alleles,
so error rate from these otherwise
not distinctive or rare loci/allele combinations of 0.7 per
cent or 1 in 140
Please inform me how many labs cross-check for this
source of error by repeating on other systems ?
False homozygosity is a systemic failing so I cannot see
how error rates can be better than 1 in 140 so sets that bound.
I come from a proper trained scientific background and
I keep seeing statements in courts where it is patently obvious
that the people speaking have no proper scientific training.
Yesterday I saw reference to 3.482 billion and last week
reference to 171 trillion. As always there was no mention of
the degree of error in these figures. My training says
that if a number is given as the result of any
processing then if no error bound is given then it is
implied by the degree of precision in the stated numbers.
The result cannot have a greater degree of precision
than the error in the most influential input factor/s.
So 3.482 billion is implied to be accurate to the nearest
2 million or even 1 million. 171 trillion is accurate to
the nearest trillion, all utter tripe.
It is just as grating to the ears of a proper scientist as
seeing someone using a calculator and pronouncing
that some result is 2.34278547 where even quoting 2.3
would be overstating the precision.
That is before bringing in the matter of the square law of
false matches once the threshold has been reached. By
the time you got to a billion profiles let alone a trillion
there would be millions of unrelated pair matches let
alone triples, quadruple matches etc. With no way of
predicting whether this defendent's profile would be
one of those false matches.
I use Identifiler and our validation and everyday use shows
absolutely no evidence of false heterozygosity.
Also, "best available" isn't always the best or even the most
useful.
False heterozygosity is not the problem - it is
false HOMOzygosity that is the irremoveable 'error'
that puts an upper bound on error rates.
I found no reference to Identifier being immune to
false homozygosities.
How often do you run a sample both through Identifier
and say Powerplex 16 which both analyse the 5
loci shown to be prone to FH. ?
I was trying to be charitable but the best in my terms would
have ben the results for Gednap 20 & 21 which show 0.7 per cent
error rate or 1 in 11 . After that result they moved the
goal posts. They dropped the stains such as "cigarette butts and mixtures
of body fluids as well as hairs". Because these samples were
producing disproportionately more errors.
Beggars belief doesn't it - just what is forensically
more likely than 50/50 blood/blood traces.
So I'll revise the worst case situation to 1 in 12 profiles
being erroneous.
If errors are so rampant how can there be any "known homozygotes"
or "known"
anythings? All must be tested by lab personnel at one point or
another -
thereby inciting more error.
How can a known homozygote give false homozygous results? Wouldn't
a known
homozygote have homozygous results?
AFAIK it is an intractable systemic failing that cannot
be overcome by normal single kit analysis.
Full explanation of biological reasons are in the full text of
which the abstract is
http://tinyurl.com/75dbu
They undertook further biological research but
full textual results/explanations hedged with
a lot of "appears"/"appeared to be"
Discrepencies on
vWA (apparently) due to mutation upstream of the repetitions
D8 " " downstream of the repeats
FGA " " downstream of repetition
D5 " " downstream of repetition
D18 mutation
Pointers to other published False Homozygosity
results for D16 and TPOX
Allelic dropout "due to stochastic effects was also eliminated"
The D5 electrophoregram is included and it is not
a matter of mis-interpretation. There is not the slightest
trace above mush level on Powerplex16
for the '10' trace of D5(10,11) and D5(10,12)
shown on Profiler +
As far as I can see this 15 in 2055 or 1 in 140 represents
the limit of error for single kit processing,
even if absolutely all human error,
mis-interprertation etc could be removed.
The 2055 subjects in the study were not pre-selected
so can be considered generally representative of
human populations.
The most simple response to false HOMOzygosity is... SO WHAT?
Missing alleles just means my stats aren't going to be as good. It
will never cause me to include someone incorrectly. So I would
dispute it as even being a relevant "error". As far as mixtures go,
I routinely see allelic dropout in my casework; at least what I
perceive as dropout based on camparison to references. Based on
your "error" any incomplete profile or mixture profiles are errors.
News flash... these are not errors. Incomplete profiles are what
they are.
False homozygosity can have some impact on databases in that the
presence of an additional allele has to be resolved in the case of a
preliminary match. However, we have not encountered this problem at
our lab. I am sure it will eventually happen.
When alleles start to "drop-in" then we have a problem. It is known
as contamination. The reverse is not true though. So you can't take
false homozygosity rates and modify statistics given the fact that
they aren't errors. The "missing" allele has already altered the
stats.
Excuse my ignorance, but could someone shed some light on 'false
homozygosity'?
Is it a missing allele in a profile? If so, would it really be labeled
homozygous?
Is it an error that makes a homozygote out of a heterozygote?
False homozygosity is when at a particular loci you only see one
allele when in fact there should be two alleles (heterozygosity). It
can result from a bad amplification, a mutation in the primer binding
site at one allele, or from a mosaic between tissues (I actually know
of someone who has this). Regardless, false homozygosity is rare.
If it is consistent due to a primer binding site mutation or mosaic
condition then the result will be present in any forensic unknown
samples and the reference. The analyst will be ignorant of
the "error". The chances of a false inclusion due to this phenomenom
would be extemely small.
So ,hypothetically, you have an electrophoregram in front of you ,it
is a profile from a single individual contributor,clean and uncontaminated
and it shows one peak corresponding to vWA 15.
No doubt about it just a single peak, perfectly on ladder.
This person , unknown to you, tested on another company's
system shows two peaks, for vWA (15,17).
By what incredible insight do you ascribe vWA (15,17)
rather than vWA(15,15)
to that single peak on your normal single system/kit analysis?
"we have not encountered this problem at our lab" -
have you even looked? , have you tried divided samples on
2 or more systems and actually compared results ?
The "so what" is it puts an unassailable bound on DNA
profile accuracy in normal single kit use.
I still have not heard properly constituted rebuttal evidence
against the upper and lower bounds for DNA profiles.
That is, until proven otherwise, between 1 in 11 and 1 in 140 for CODIS
and 1 in 14 to 1 in 140 for FSS/NDNAD.
My reply is that I don't care. It can happen. I can't change it. I
would love to see the RFU levels of the profile in which the 17 was
absent. The bottom line is that this occurance is rare and as long
as the 17 is absent in every sample then it doesn't effect my
interpretation of a match.
The FSI 143 (2004) 47-52 report only has plots for the D5
'11,11' and '12,12' Powerplex plots that showed (10,11)
and (10,12) on Profile, no RFU levels .
Citations in that article give further FH results by other
workers so not one isolated study :
J. For. Sc. 46 (2001) p637-641 also shows FH of 0.2 per cent for FGA
" " " 43 (1998) 1103-1104 shows D16 and TPOX FH
http://journalsip.astm.org/JOURNALS/FORENSIC/jofs_toclist.html
shows vol/year agreement to these refs but I cannot find
the cited articles/abstracts there.
You have not given rebuttal evidence , just a
qualative counter assessment. 3 quantitive studies to me, 0 studies
to you is not rebuttal. Where is your data for the upper and lower
error bounds for single-kit DNA profiles ?
Qualatative is all that is needed and there are no error bounds for
the kits that I know of.
As far as I am aware
Profiler, Cofiler, Identifiler, GenePrint and PowerPlex
are all compliant with the CODIS requirements
and results submitted for storage on Codis. These are not
mutually consistent as far as false homozygosity (FH) is concerned.
It is outrageous that the FH inconsitencies , hence
intractable systemic errors, are
not determined published and presented in courts.
On the more general theme of errors (human included)
I now see that the latest published Gednap report is on the web
rather than locked away as expensive subscription access in
Int J Leg Med (2004) 111:83-83
http://medweb.uni-muenster.de/institute/remed/spurenkommission/Information/IJLM_GEDNAP_II.pdf
Gednap 24/25 fudged to bring the per locus error rate down from the 0.7 per
cent of the earlier Gednap 20 validation trial
GEDNAP 20 the following samples wereprepared:GEDNAP 201.
Person A: 25µl blood (female) on cotton cloth2.
Person B: 25µl blood (male) on cotton cloth3.
Person C: 25µl blood (male) cotton cloth
4. Stain 1: unsmoked filter cigarette with 10µl saliva
5. Stain 2: 25µl blood mixture (Persons A:B, mixed 1:2 v/v)
6. Stain 3: 25µl blood mixture (Persons A:C, mixed 3:1 v/v)
7. Stain 4: Buccal swab from Person A
...
Being the previous criteria, now downgraded to 50/50
blood mixed stains only, no cigarettes etc, to bring the per locus error
rate down from 0.7 to 0.4.
Despite being a "blind trial" the stains were mixtures of
persons A,B and C, known, in these outline terms,
to the participants. The full protocols for Gednap 24 to 27 are
not available to the public AFAIK (rat smelling time -
cf concurrent thread on this group about "ASCLD-LAB accreditation" ).
They were rigorous enough to include THO 9.3 & 10
and FGA 42.2 (perceived as trip-up ones by some )
but if I was running the tests I would make sure a
few False Homozygous prone alleles were included eg
from pool of the vWA,D8,FGA,D18 or D5 ones mentioned in
FSI 134 (2004) 47-52
