Sciencemadness Discussion Board

False DNA-test-positives possible ?

chief - 3-10-2008 at 04:11

Everyone knows it: From time to time another crime is investigated via DNA-sequencing loads of samples, sometimes maybe 10000's of samples, from the entire population of an area.

That rises the question: How secure is this method really ??? If there is a possibility for false positives, then the unfortunate one from who the sample was taken will have the outermost difficulty defending himself ...
And in the media the DNA-sequencing is seen as 100 % foolproof ...

Since it's a chromatographic method (as far as I know) it should be relyable most of the time, but then there is eg. an example from Germany, where it seems that the method might have some yet unknown difficulties:
http://images.zeit.de/text/2008/18/Die-Unsichtbare
.. after all the human genetics
==> is not very far explored
==> the method is young
==> mistakes from _that_ busines usually are kept under closure and never surface, or are not even recognized as such ...
==> ...

So: What examles might there be to view this matter in an investigative and critical light ??

not_important - 3-10-2008 at 05:40

There are several different methods used for DNA analysis. The older RFLP analysis has a fairly high probability of mismatch, perhaps as high as 1 in 1000. The newer methods have much lower probabilities of misidentification.

When used as proof of an individual not matching the sample, the tests are quite reliable. When used as proof they do match, the reliability goes down from several factors. If a group supplied to evidence DNA, say from the victim of gang rape, then matching any given individual is being done against the union of all the perpetrators which increases the odds of false positives. In effect a suspect is being tested to see if they have a given VNTR matching any one of the perpetrators, a match can be found that (unknown to the tester) comes from perpetrator A, another VNTR matches that of perpetrator B, and so on. This increases the chance of a false positive.

Other factors are contamination of samples in gather them or in the laboratory, poor QC in the lab (automation is reducing this), and restricted gene pools as found in isolated groups.

For those reasons a DNA match is much less useful as proof, than as negative proof; it is good at showing someone is not the person who originated the DNA sample but less certain when being used to prove that they are the source of the evidence DNA.

Some places do not allow positive DNA matches to be used as evidence, or as the major or sole evidence, for conviction. A good defense attorney will know enough to correctly present the difficulties with positive matches, while the prosecutor often misstates the tests uniqueness. Such misunderstanding of statistics has long been a problem with prosecution and goes back well before DNA testing, there was a case in the US some 40 or 50 years ago where a couple was convicted based on misuse of statistics - the odds of a couple of the same physical description being in a automobile of the given make and model were incorrectly combined to give 'odds' as 1 in millions rather than 1 in a few hundred. In that case the actual guilty couple was later found, but prosecutors continue to be ignorant or willing misuse statistics.

chief - 3-10-2008 at 06:53

Thats interesting:
==> When in some places it's not allowed to use the DNA-sequencing as main proof or as proof at all,
==> then there they will have written down, officially, the reasons for that (!),
===> because usually one group (law enforcement) proposes for the use of such new methods,
===> then it's reviewed by experts,
===> and they give their reasoning to the decision-makers ...,
so it probably is documented somewhere, even with examples ... !

chemoleo - 3-10-2008 at 18:13

Ultimately, sequencing of key genes (plural) (rather than RFLP) will determine the identity of the DNA sample. I.e. there's a certain allelic and somatic variability of genes upon which restriction fragment length polymorphism (RFLP) is based, so the more genes are sequenced, the less is the likelihood for misidentification, i.e. false positives. RFLPs are very crude compared to gene sequencing as they only recognise the variability of restriction enzyme sites, but not the entire gene/chromosomal sequence. I'm sure that in the not-too-far future gene sequencing will be used instead, eliminating false positives entirely. Unless you are a twin.

Just to update you, DNA sequencing technology is so good that it is possible to sequence Neanderthal or even Mammoth genomes. Genomes, that is, not genes, so the entire gene content of these 'organisms'. Sequencing a sperm or saliva sample by comparison is easy.

The fallacy with DNA sequencing and fingerprints is the same - a fingerprint or DNA sample (hair, drop of spit, skin abrasion etc) can be planted to make investigators think that you were there. But I'm sure investigators are aware of that.

Panache - 7-10-2008 at 16:12

Quote:
Originally posted by chemoleo

The fallacy with DNA sequencing and fingerprints is the same - a fingerprint or DNA sample (hair, drop of spit, skin abrasion etc) can be planted to make investigators think that you were there. But I'm sure investigators are aware of that.


The greater fallacy of course being the supposition that no two humans (that are not identical twins) have the same DNA or fingerprints. You could not confirm this unless all humans were directly compared. Law enforcement has for years used the simple marketing speak to imply certainity in these tests, whereas in fact no one can correctly offer certainty.
Anyone with any experience with expert witness work can attest to the level of coaching 'your side' will attempt with you to temper your language. I refuse to use or agree to phrases using the word 'impossible' preferring instead 'improbable', but the difference in effectiveness is marked. Luckily most civil matters involving large amounts of technical 'opinion' in Australia are heard and decided upon by judges only, limiting the effect of bullshit.

[Edited on 7-10-2008 by Panache]

chemoleo - 7-10-2008 at 16:43

Quote:
he greater fallacy of course being the supposition that no two humans (that are not identical twins) have the same DNA or fingerprints.


Well, not so, this is undoubtedly 99.99999999999....% true, at least regarding the total DNA.
There are 3 billion bases in the human genome, and four of each type - so there are 4 ^ (3,000,000,000) combinations... decidedly a very very large number!

DNA fingerprinting is a different issue, there it is very unlikely but remotely possible that someone with an identical (restriction enzyme fragment length polymorphism) pattern is found. After all fingerprinting only looks at a very small sample of the genome.

Twospoons - 7-10-2008 at 18:38

There may be 4^(3,000,000,000) combinations, but how many of those result in a functional human body? Probably only 100 billion or so. Another example of the misuse of statistics.

chemoleo - 7-10-2008 at 19:12

Lol thank you, very good.
I was wondering whether someone would spot it, but was hoping for a keen reader.

It is similar in principle to the Levinthal paradoxon - to quote:
Quote:
In 1969 Cyrus Levinthal noted that, because of the very large number of degrees of freedom in an unfolded polypeptide chain, the molecule has an astronomical number of possible conformations. (The estimate 3300 or 10143 appears in the original article (for a discussion see [1]). If the protein is to attain its correctly folded configuration by sequentially sampling all the possible conformations, it would require a time longer than the age of the universe to arrive at its correct native conformation. This is true even if conformations are sampled at rapid (nanosecond or picosecond) rates.


The fallacy here is, many of the peptide conformations would clash in 3D space, so are not possible. Therefore the number of possibilities of protein folding are severely limited by many order of magnitudes.

Same thing with the genome - many mutations would result in embryonic death, or (in those comparably rare cases that evolved to) other organisms entirely, so what we can observe is already selected for what represents our species and what can survive.

However even (and this is a very very big even) if each human only differed by one mutation from another, there would be 3 billion possibilities. And we all know, the difference from one human is much greater than that, there are many (hundreds?) of thousands of differences - the differences that make us individual.

Which leaves us again with an astronomically large number of viable human genotypes, way greater than the mere 6 point something billion the human race currently holds.

Panache - 8-10-2008 at 04:40

Quote:
Originally posted by chemoleo

However even (and this is a very very big even) if each human only differed by one mutation from another, there would be 3 billion possibilities. And we all know, the difference from one human is much greater than that, there are many (hundreds?) of thousands of differences - the differences that make us individual.

Which leaves us again with an astronomically large number of viable human genotypes, way greater than the mere 6 point something billion the human race currently holds.


Right therefore it is only improbable!

OMG WTF PWNT!!!


:D

Jdurg - 6-11-2008 at 17:34

Well here in the USA DNA evidence alone can NOT result in a conviction. It is not considered "primary" evidence. It is secondary evidence. When used with enough other evidence that links a suspect to the crime, the DNA can give the last bit of information needed to convict them.