Wednesday, March 03, 2010

Small, but perfectly formed – judicial acceptance of LCN DNA analysis

Updating the forensic use of LCN DNA analysis: R v Wallace [2010] NZCA 46 surveys developments and refers to R v Reed & Garmson (2009) EWCA Crim 2698. Brief mention of Bayes' Theorem, led the Court in Wallace to the following "insight" (para 87):

"As a matter of logic, what all this means for criminal trials is that with low level DNA profiling it is not so much a question of the particular technique or method. This can, if rather painfully, be explained to judges and jurors. It is the philosophy of the approach that is important: the uncertainty of the method of transfer does not diminish the fact of the matching DNA profile for the former does not invalidate the latter. The method of transfer however is of great moment for the court, whilst it does not necessarily concern the scientist."

That is to say, a match is a match is a match, but the court needs to know why there is a match.

The Reed & Garmson dicta were summarised in Wallace (para 97):

"Three points can be made about this decision. First, it is apparent that the Court of Appeal of England and Wales accepted that at least under defined conditions Low Template DNA can form admissible profiles, capable of reliable interpretation. Indeed, during the course of the Reed & Garmson appeals both appellants abandoned their general attack on the reliability of Low Template evidence using the LCN process. Second, the Court of Appeal of England and Wales prescribed what might be called a "safety" floor limit. Above the range described by that Court as being a stochastic threshold of 100 and 200 picograms results can be so produced which are capable of reliable interpretation. Third, in the view of that Court a challenge to the validity of the method of analysing DNA by the LCN process should no longer be permitted where the threshold limit is exceeded." [the Court's emphases]

There may be some circumstances in which whether there is a match can be an issue, but these are limited (on current understanding) to cases where the amount of DNA from the crime scene was so small – around 100 to 200 picograms (1 picogram = a millionth of a millionth of a gram) that random errors (called stochastic effects) could cast doubt on whether a match existed (applying Reed & Garmson).

In Wallace it was argued for the appellant that it was incorrect to regard the relevant sample of DNA from the scene as coming from a mixed source, and therefore it was wrong to conclude that it did not exclude the accused. This was an issue of interpretation and the evidence supporting the appellant's view was not in this case sufficient to cast doubt on the correctness of the verdict. The Court emphasised that DNA evidence must always be evaluated in the context of the other evidence in the case. In Wallace the DNA evidence which was the focus of this appeal was not significant in that context.

I should add that I have summarised the facts in Wallace a little inaccurately in the interests of simplicity. The DNA just referred to was actually alleged to have been from the victim, and was found on a metal bar, allegedly the murder weapon, which the prosecution claimed to link to the accused.

Another sample of DNA was from the victim’s shoelace, and was allegedly from the accused. However this latter DNA was (para 33, 38) thirteen times more likely to have come from the accused or a paternal relative as from any randomly chosen New Zealand male. The trial court, as emphasised by the Court of Appeal, made it clear to the jury that this evidence did not prove the identity of the accused as the murderer, but it was merely consistent with the prosecution case. In reality – and contrary to the prosecution’s disavowal at trial - this DNA evidence was highly probative of guilt.

To say that the DNA was thirteen times more likely to have come from the accused (or a paternal relative) than from a randomly selected male means that if the likelihood of a random match is x, the likelihood of a match with the accused or such relative is 13x. The Bayes likelihood ratio is 13. Despite the protestations to the contrary at trial, this evidence was highly probative of guilt.

The appellate court needed to be satisfied that the other evidence in the case proved beyond reasonable doubt that the accused was the murderer. This is because, in Baysean terms, if the prior probability of guilt was only 0.90 (let’s assume this is not proof beyond reasonable doubt), the DNA on the victim’s shoelace would have increased that to a probability of guilt of 0.99. That is, this DNA evidence would have removed a reasonable doubt if the other evidence only proved guilt to a probability of 0.90. It would have been highly probative and important to the prosecution case, and errors in its interpretation would have been significant.  

No comments: