One of the two exhibits in the trial of Amanda Knox and Raffaele Sollecito to be looked at by Stefano Conti and Carla Vecchiotti as part of their review of DNA evidence in the case is a knife, recovered from Sollecito’s flat and alleged by prosecutors to have been used by Knox to stab Meredith Kercher in the neck.
Conti and Vecchotti agree with the opinion of the chief police scientist, Patrizia Stefanoni, that Knox’s DNA was present on the handle of the knife and they don’t dispute the match between the DNA found on the blade and that of the victim. But they raise clear concerns that the DNA profile observed in relation to the blade could have been the result of contamination.
The reasoning behind this turns on their characterisation of the the DNA sample from the blade as “Low Copy Number” (LCN), essentially meaning that there were so few examples of the DNA in question within the sample that, according to mainstream scientific opinion, special measures would have been required so as to ensure the integrity of the testing.
In her trial testimony, Stefanoni appeared to accept the proposition that the sample in question may indeed have been an LCN sample. It appears difficult to be absolutely sure about this, because there is no accurate record of the quantity of originating material (Stefanoni has said it was “in the order of a few hundred picograms (pg)”). Quantifications during a process known as qPCR, it seems, would have given a strong indication, but there is no record of these having been performed – perhaps because of the very fact that the sample was so small as to be out of the range of sensitivity offered by the equipment that would have been used.
However, the review highlights two reasons to suppose that the sample must have been of an LCN-type quantity. Firstly, there are what are known as “peak imbalances” in the graphical representations of the DNA which were produced. In most graphical printouts of DNA tests, peaks come in pairs. In an imaginary perfect test, the pairs would all be of equal height to one another. That is rarely the case in practice, so a ratio of 3:5 is normally considered acceptable. But, it can be seen from the graphs reproduced from page 68 on in the review that some of the pairings are more imbalanced than that. Although these imbalances can be produced by a number of means, they can be indicative of the presence of LCN material, so this provides some support for the notion that we are dealing with LCN. Moreover, if you accept that the raw material was “a few hundred picograms” and note that the volume of the solution from which the DNA reading was obtained was 10 microlitres (μl), then it follows that the concentration of that solutions must have been a few tens of pg/μl. According to the review, 200 pg/μl is generally considered to be a threshold below which material should be treated as LCN.
One difficulty with analysing an LCN sample is that, because its presence in a solution is so faint, amplification is required during the PCR process (indeed, it seems clear that Stefanoni did carry out amplification on the sample). Think about this as being like turning up the volume on a cassette player to hear something that has been recorded at a very low volume. Inevitably you are also turning up the hiss. Background noise will normally be easy enough to distinguish from the genuine data, but the real problem is that, as you amplify, you may be amplifying faint traces of contamination that are also present. So, in our case, if there is a reasonable risk that Meredith Kercher’s DNA might have been present as a trace-level contaminant, then the test conducted ought not to be considered reliable.
Conti and Vecchiotti portray this a real risk that could have arisen at any stage of the investigative process, from the recovery of the knife through to the final tests in the lab. They point to a body of scientific literature recommended specific procedures, above what may typically be implemented, in order to minimise the risk of contamination in the recovery, handling and analysis of the evidence, and question whether such procedures were applied in the Kercher case. It is even suggested that the sample from the blade should have been tested in an entirely different laboratory.
I think the strength of this argument varies according to the context you are talking about. Failure to protect against contamination is not necessarily to be excused at any stage. But, from the point-of-view of the results we are considering, it only really matters if there is an appreciable risk that the data observed could have resulted from contamination. So, it should be considered that the location from which the knife was recovered – Sollecito’s flat – is not one where the victim’s DNA would be expected to be present, because it is somewhere where she had never set foot. Regardless of how suitable the procedures carried out there were, there does not seem to be a real risk of contamination.
Similarly, in between the flat and the lab, the knife does not appear to have been in any location where Kercher’s DNA might be supposed to have been present nor handled by anyone who had been in contact with Kercher’s DNA.
As a side note, the review suggests that secondary contamination can never be ruled out in DNA testing, points to an authority that detecting this type of contamination is more likely where LCN is concerned. For example, perhaps Sollecito shook hands with Kercher at some point, then carried her DNA back to his flat and contaminated the knife. In theory, it is probably correct to say that this possibility cannot be ruled out. But the prosecution may be expected to argue that this type of contamination is an ever-present and merely theoretical risk. What the court ought to want to ascertain is exactly how likely such contamination is.
Even if it might be argued that contamination from Kercher’s DNA at Sollecito’s flat or during the initial handling of the evidence is a remote possibility, it is certain that her DNA was present in the lab. At least 50 samples containing it were tested there, 20 or more of them prior to the testing of the knife.
For this reason, a key question for the court is whether the procedures in the lab were adequate for minimising the risk of cross-contamination of samples by the people working there, particularly given the apparent likelihood that the material in question was an LCN sample. The authors of the review don’t categorically say that they weren’t, but that they don’t have enough information to make a judgement. The court at the original Knox/Sollecito trial was satisfied that the standards of the laboratory were adequate in this respect. Will the appeal court take the same view? Will it feel obliged to hold the police scientists to higher standards in light of the review?
If the procedures in place (an actually implemented) in the lab are found to satisfactory, that is not necessarily all there is to it. There are checks on contamination that can be carried out but were not in this case, perhaps in part owing to the small quantity of material being studied.
Ideally, as was conceded by Stefanoni in the original trial, the sample should have been broken into two or more portions to be tested separately. Assuming that similar results would have been replicated on at least two occasions, this would have shown contamination to be less likely.
As a further precaution, the presence of some type of organic material might have been established. Now, it may be felt that the victim’s DNA is the victims DNA. What does it matter whether it came from blood, or skin or whatever? However, if the presence of some particular type of cell were known, then this would provide an indication against the idea that the results were contamination. Stefanoni seems to have performed a test for blood which was negative, and then proceeded on the basis that the sample was too small for the test to have worked properly.
It seems likely that dividing the sample simply wasn’t possible and, although Stefanoni may now wish that she had been more thorough in determining the biological nature of the material she was dealing with, that lapse is probably not enough in itself to invalidate her results.
A bigger problem to my understanding (although the review doesn’t stress it in the way I might have expected) is the issue of how control tests were performed. Standard procedure is to run negative controls, where the test is performed on a saline solution or something else not expected to contain any DNA, so as to check for the presence of contamination already present in the equipment or tube. A positive control should also be run, on a sample whose properties are known, in order to check that the equipment is behaving as it would be expected to (a bit like checking your bathroom scales using a bag of sugar).
These control tests should really be performed not only at the start of testing for each sample, but also when an amplification is performed on a sample. This is particularly important for the negative control, because of the possibility that the amplification will reveal contamination that may not be evident prior to the amplification. But Conti and Vecchiotti report that they see no evidence that this was done. Perhaps the bench will be satisfied if Stefanoni performed the controls without saving them electronically and is able to confirm that no problems were observed. But there’s a clear problem here for the prosecution if it emerges that the controls were not performed.
The most important things for discussion at the next hearing on 25 July will, I think, be the quality of the procedures in place in the police lab for reducing the risk of cross-contamination between samples and the issue of whether or not adequate controls were performed on the knife-blade sample. The reliability of the knife as a piece of evidence in the case really relies on those two things going in the favour of the prosecution.