Thursday, March 17, 2011

An analysis of the bra clasp DNA

Part 27 in the Knox/Sollecito case

Update (7 July 2012)

Some time ago I reviewed some passages in the English translation of the Massei report that concern the bra clasp DNA and examined better electropherograms.  I would like to discuss these without revealing the reference profiles of any of the individuals.  The interested reader should also consult the translation of the Conti-Vecchiotti report and this blog.

Stutter peaks
On page 206 Stefanoni used the word “noise” in a different manner from the way a spectroscopist would use it. From page 207, “Thus, where there is an allele which has a certain height and such that the peak just before it has a much smaller height, at most 15% of the first one, then the previous peak should be considered noise, just a by-product of the analysis.” One infers that Stefanoni used the word noise to refer to stutter peaks and possibly to refer to other artifacts such as blobs. Tagliabracci also used the word noise to mean stutter on p. 241.

If an allele is found at 17 repeats, one expects to see a peak at 16 repeats that is less than 15% the intensity of the peak at 15 but rarely does one see a stutter peak at 18 repeats.  The basis of stutter is that the primer and the template DNA strands do not always anneal perfectly in the DNA replication process. It is far more likely that the bulge of one extra repeat unit will occur on the template strand than on the primer strand, and that is why the stutter peak almost always has one repeat unit less than the true allele.

A problem for a forensic DNA scientist is that when a mixture consists of one strong and one weak profile, discriminating between the weak profile and the stutter peaks is very challenging.  John Butler (p. 125) wrote, “Mixture interpretation requires a good understanding of the behavior of stutter products in single source samples.”

Threshold values
On page 207 one reads, “The height which is considered reliable for a peak to be qualified as an allele is equal to 50 RFU, the symbol RFU representing the unit of measure employed for these measurements.” This is a remarkable statement in one respect, inasmuch as if Stefanoni actually adhered to it, about 22 of about 29 peaks attributed to Meredith on the knife profile would fail to be scored. In other words, she did not respect her own minimum threshold value in at least one other experiment.

On pages 208-209, one encounters some strange statements. “She was asked if she had considered that peak, number 13 [This peak is in locus D5S818 and its height is 108 RFU], as an allele or as noise. Dr Stefanoni declared that she had not considered that peak as an allele or as noise… it can't be an allele because it is too low with respect to the main peaks.” This argument does not make sense; either a peak is an artifact or it is real. It is not in the correct position to be stutter, and there is no reason why a true allele can’t be smaller than Meredith’s profile. Moreover Tagliabracci questioned Stefanoni’s interpretation of this allele (pp. 241-242), noting that in locus vWA that she had taken a peak of only 65 RFU as an allele. Stefanoni argued that each peak should be judged on a case-by-case basis (p. 209), including information such as the main peak heights. Yet the main peaks in vWA are 84% as high as the main peaks in D5S818 on average, whereas 65 RFU is only 60% of 108 RFU. In other words what objective criterion Dr. Stefanoni used to reject a peak of 108 RFU and keep a peak of 65 RFU is obscure or nonexistent. Finally, there are other peaks in vWA at 17 and 18 repeats that are not labeled yet the larger of the two is about 50 RFU. It is difficult to see why this peak was not considered an allele.

Stefanoni was asked about an alternate interpretation, one in which a minor contributor would have the alleles 12 and 13. “In response to this question-observation, Dr Stefanoni explained that in this case, it would not be possible to explain the Y chromosome, and thus reaffirmed the correctness of the interpretation she had given.” It sounds as if Stefanoni used her attribution of the Y-chromosome profile to Raffaele to then interpret the autosomal DNA as having his profile.

Locus D21S11 is problematic (pp. 241-242). Stefanoni counted as stutter a peak that is 15.6% the height of the next peak. That is higher than the 15% cutoff, which is higher than stutter with which I am familiar. But she counted as real a peak that is 17.2% of the next peak, which is not that much of a difference. This peak could constitute half of Raffaele’s profile in this locus, but the other half would fall underneath one of Meredith’s two peaks, which are roughly sixfold higher. Although it is possible that Raffaele’s other allele is present and contributes additional intensity to the second of the two large peaks belonging to Meredith, I don’t see any reason to assume that it must.

There are peaks on the bra clasp electropherogram that are not part of Raffaele’s profile and are not stutter. In reviewing the bra clasp DNA, I am more concerned than I was before that Stefanoni applied a suspect-centered approach, which would be contrary to good forensic practice.

A previous post has examined the DNA on the bra clasp, which has a strong profile from the victim, Meredith Kercher. In this post we will examine the bra clasp DNA on the basis of the available information with respect to whether or not Raffaele’s DNA is present. We will also look into the uncertainties surrounding whether or not the DNA of a third party’s DNA is present. However, a complete analysis would require electronic data files, not electropherogram images, and the defense never received these or other files. All references to the Massei Motivations report are given with respect to the English translation available at Dr. Adriano Tagliabracci was an expert witness for Raffaele Sollecito’s defense, and his interpretation of the bra clasp DNA profile clashed with Dr. Patrizia Stefanoni’s, the chief prosecution witness with respect to DNA forensics. The bra clasp also figured prominently in an open letter coauthored by two DNA forensic experts and cosigned by seven others.

Executive Summary
Y-chromosomal DNA corresponding to Raffaele’s haplotype was found on the clasp as well as many alleles corresponding to his autosomal DNA. Dr. Tagliabracci’s critique of Dr. Stefanoni’s analysis, that it was suspect-centered, is probably valid but is tangential to the more important questions surrounding the collection of the clasp. The reanalysis of the clasp in the appeal will likely conclude that a partial profile corresponding to Raffaele’s autosomal DNA (but possibly not a full profile) is present. However, there are one or more other contributors as well, apart from Meredith, and the more salient question is how this DNA was deposited on the clasp. There is no reason to suppose that the third person’s DNA necessarily arrived via a different mechanism from Raffaele’s: Secondary transfer, contamination at the crime scene, and evidence-tampering are all more likely than Raffaele’s depositing DNA on this dubious piece of evidence during the murder, but not depositing DNA on the bra itself or leaving any other trace in Meredith’s bedroom.

Analyzing Mixtures of DNA
It is worth reiterating that analyses of DNA mixtures are somewhat subjective. With respect to the John Puckett case, Chris Smith reported: “Mixed samples are another flashpoint in the DNA wars. It can be exceedingly difficult to separate one person’s DNA from another’s, especially in a degraded sample like this, and there is no universally accepted way to interpret the resulting profile. As the eminent British researcher Peter Gill told a conference of his fellow forensic scientists in 2005, ‘If you show 10 colleagues a mixture, you will probably end up with 10 different answers.’ Even Cheng’s supervisor, a combative man named Matt Gabriel, reluctantly admits on the stand that there is no agreed-on protocol for handling mixed samples.” With respect to the Puckett case Michael Bobelaian quoted Professor Dan Krane: “’There is a public perception that DNA profiles are black and white,’ he told me. ‘The reality is that easily in half of all cases—namely, those where the samples are mixed or degraded—there is the potential for subjectivity.’”

An article in the Journal of Forensic Sciences provides good discussion of how to minimize observer effects in DNA forensics. Among the authors are co-signers of the Johnson-Hampikian open letter and one person who has written extensively on the problem of investigator bias. Note that the proposal indicates that the investigator should not have access to the reference samples until quite late into the analysis process. In other words, one wrong way to analyze an evidence sample would be to lay reference sample electropherograms on top of it and look for matches. Overlaying the reference sample onto the evidence electropherogram after the analysis is complete simply to present the data is fine.

Stutter peaks
Individuals vary in the number of short terminal repeats of DNA at various locations in their chromosomes, and this is the basis of DNA profiling. Larger fragments of DNA emerge later from the capillary tube and give rise to peaks in an electropherogram. Each peak corresponds with one allele. Stutter peaks are a type of artifact in the electropherogram, and they most often show up at a position on the time-axis (which is proportional to the size of the DNA fragment) that is one repeat unit shorter than the true allele. They are usually in the vicinity of 5% of the height of the true peak with which they are related. For the purposes of this analysis, we will assume that any peak that is one repeat unit shorter than an allele in Meredith’s profile is likely to be a stutter peak.

From the Massei Report, about page 243: "Regarding locus D7S820, he [Dr. Tagliabracci] revealed that Forensics had interpreted it, recognizing the presence of two alleles, 8 and 11; they had not taken into consideration a peak, low, but still higher than 50 RFU, corresponding to allele 10." To whom does it belong? In locus D16S539, Dr. Tagliabracci believes that there is a peak at the 13 locus. This allele is also not part of Meredith’s or Raffaele’s profile, but it is quite possible that both are stutter peaks.

Other DNA on the clasp
However, there are other small peaks besides the one at 10 in the D7S820 locus, and because of their positions along the time-axis, it is difficult to believe that they are stutters. In locus D19S433 there are alleles at 10 and 14 repeats that are not part of Meredith’s or Raffaele’s profile. In locus vWA there appear to be peaks at 9, 11, and 18, that are not part of these individual’s DNA profiles. In other words, there is probably DNA on the clasp that belongs to someone other than Meredith Kercher or Raffaele Sollecito. Although it is difficult to say how many individuals contributed, there would seem to be at least two. The critical question is how it came to be there, taken up below. Some have tried to claim that Amanda was a third contributor to the bra clasp DNA. However, there can be no doubt that Amanda's complete profile is not on the bra clasp.

The presence of these small peaks raises another question. Many of these small peaks were not marked on the electropherogram of the clasp, suggesting the possibility that Dr. Stefanoni’s lab did not consider them to be real peaks. Some of them appear to be in the range of 100 RFUs. Although they are small relative to typical peaks, they are as large or larger than all of the peaks in Meredith’s profile from the kitchen knife. This forces one to ask whether a consistent peak threshold were used for all samples, and if not, one has to question whether peak thresholds were changed ex post facto, something that seems contrary to basic principles.

Raffaele’s putative DNA on the bra clasp
Raffaele and Meredith shared eleven alleles out of about thirty. Because Meredith’s profile is roughly eightfold more intense than Raffaele’s putative profile, it is very challenging at the very least to know whether Meredith’s DNA alone or Meredith’s plus Raffaele’s DNA contributed to the strong (>1000 RFU) peaks in the profile. For one thing, peaks heights can decrease from left to right across an elecropherogram due to degradation, which attenuates peak heights corresponding to large DNA fragments more than smaller loci. For another, small peaks vary in intensity more than large ones due to noise and stochastic effects. An additional complication is that Raffaele’s profile has a few peaks that would show up at the same position as stutter peaks from Meredith’s profile would.

Let us now turn to the alleles in which Raffaele’s reference profile is distinct from Meredith’s profile. “Professor Tagliabracci then maintained that this suspect-centric method was detectible in Dr. Stefanoni’s report and presentation because, he affirmed, it was a case of forcing the profile obtained … eliminating or leaving out alleles [257] solely for the purpose of making that profile compatible with Raffaele Sollecito’s profile.” (Massei Report translation, p. 241)

The presence of Raffaele’s DNA in YSTR testing strengthens the case that Raffaele’s autosomal DNA is present on the clasp, though in low quantity. One way to take some of Dr. Tagliabracci’s objections into account is to acknowledge that Raffaele’s DNA is present in the low template number range, as discussed in the appendix below.

Some problems in the collection techniques
As noted above the key question is how did Raffaele’s DNA and the DNA of one or more other individuals become deposited onto the clasp. Contamination can happen when an item of evidence is collected at the scene of a crime, as well as when it is tested in the laboratory. In this series of photographs the same crease in a glove of a forensics worker is seen, indicating that this glove was not changed. On page 13 of the book “Angel Face” Barbie Nadeau reported on Dr. Stefanoni’s cross-examination by Raffaele’s lawyer Giulia Bongiorno. Ms. Bongiorno noted that Dr. Stefanoni’s bracelet was seen in the same position above her glove, again pointing to gloves not being changed.

Dr. Stefanoni’s view was that contamination was much less likely with dry traces than with liquids (Massei Report translation, pp. 201-202; 204-205). “Regarding the possibility of transferring exfoliated cells that may be found on a hand or a glove, Dr Stefanoni explained that in the abstract, anything could be transferred, but it remained to be seen in practice. So, with specific reference to exfoliated cells, she stressed that it would be necessary to press down with force or scratch over a surface where these would have to be present (for example, the back of a person)… With reference to the single-use gloves, Dr. Stefanoni specified that they were changed, in the course of the search, every time an object was touched that was particularly soaked with blood, and when it was obvious that the gloves would be soiled; ‘otherwise, if it is just an ordinary object…I can move it, but this does not lead to my DNA remaining, let’s say, attached. It depends on the object.’” (Massei Report translation, pp. 202-203). One infers from her comments that in the absence of blood or obvious dirt or grime, gloves were not changed. Dr. Stefanoni’s views on how frequently gloves should be changed are not shared by any forensic scientists that I can identify. On page 38 of “Forensic DNA Typing,” the most authoritative textbook on the subject of DNA profiling, John Butler wrote, “Use clean latex gloves for collecting each item of evidence. Gloves should be changed between handling of different items of evidence.” Other guidelines make the same or similar recommendations.

“She [Dr. Stefanoni] confirmed, therefore, that before having touched the clasp with those gloves, the gloves had not touched any other objects, since they had just been put on.” Yet some of the gloves were dirty, including one used to handle the bra clasp. The clasp was handled by at least two people wearing gloves when it should have been handled by one person using a disposable pair of tweezers. The bra clasp was also recovered more than a meter from where it was originally observed and noticeably dusty. In summary the late-collection and subsequent handling of the clasp substantially weaken its evidentiary value, as noted in the Johnson-Hampikian open letter of 19 November 2009.

Dr. Stefanoni’s view that dehydrated traces are very hard to contaminate is also open to question. In the case of Gregory Turner previously discussed on this blog, a forensic worker contaminated fingernail clippings with both her DNA and Mr. Turner’s DNA from a ring, and there is no reason to believe that liquids were involved. Overall, the techniques in Dr. Stefanoni’s laboratory were not as stringent as they might have been, and this raises the chances of contamination, all other things being equal.

The presence of DNA from one or more persons who are not suspects on the clasp is one of the most serious black marks against the bra clasp as evidence of Mr. Sollecito’s involvement in this crime. It is difficult to see how one or more unknown individuals deposited DNA by primary transfer; therefore, secondary transfer and contamination need to be considered carefully as mechanisms by which both Raffaele’s DNA and the DNA of one or more unknown individuals arrived on the clasp.

One way to interpret Dr. Tagliabracci’s remarks is that he criticized Dr. Stefanoni for focusing on peaks in the clasp profile that happened to be in Raffaele's reference profile and ignoring peaks that did not. If Dr. Stefanoni did so because she had prior knowledge of Raffaele’s profile, then such analysis is open to question, as indicated above. Dr. Tagliabracci implied that this was a problem. “He pointed out that that there is a significant subjective element in reading the electropherograms. He focused in particular on locus D5S818, in which two principal alleles are present; together with a third peak with a height of 108 RFU; as this is higher than 50 RFU, it should have been considered an allele. Forensics [la Polizia Scientifica] did not, however, consider this to be the case; instead, they considered the 65 RFU peak to be an allele and observed that, in this way, a compatibility with Raffaele Sollecito’s profile resulted, which otherwise would not have been the case (page 59). With reference to this, Professor Tagliabracci repeated that there was a forced interpretation, which was typical of a suspect-centric attitude (page 60).” (Massei Report translation, p. 242) The quotation above only provides the peak height in RFU, not the number of repeats, which sets the location along the horizontal (time) axis. I do not know which peaks are meant.

Dr. Tagliabracci’s approach seems to be more in line with the spirit of recommendations in the 2008 Journal of Forensic Sciences article cited above, although Dr. Stefanoni denied that she used a suspect-centered approach. The Massei report gives at least five instances where Dr. Tagliabracci differed with Dr. Stefanoni with respect to the interpretation of certain loci. Specifically with respect to D5S818, he challenged her interpretation when she took a smaller peak (65 RFU) in preference to a larger one (108 RFU) to be part of a profile. Because other peaks attributed to Raffaele are generally larger, in the range of 200 RFU, this objection needs to be answered.

The Y-STR testing strengthens the case that Raffaele is indeed a contributor to the bra clasp DNA. Yet how can one explain the inequalities in peak heights in the peaks that correspond with his profile? It is possible for DNA belonging to Raffaele to have highly unequal peak heights for several reasons. One possibility is that this would happen when the non-Meredith DNA is in the low template number range. This could lead to large disparities in peak heights in the two alleles within any single locus because of stochastic effects. Dr. Tagliabracci was aware of this issue, as noted in the Massei Report translation, p. 240. Raffaele's putative DNA on the clasp is very weak, less than 200 pg according to his appeal. The exact amount of DNA depends on the details of the calculation. If the low-template number explanation is invoked to explain differences in peak heights, then the standard protocol for dealing with low template DNA is to run the sample at least twice. A retest has yet to be done but may become part of the appeals process.