Activity vs. Source Level Propositions: It Looks like Raffaele Sollecito's DNA Was on the Bra Clasp, But How Did It Get There?

Was Italian Judge Claudio Pratillo Hellman's refusal to order more DNA testing of a knife in the Knox-Sollecito case "bad judicial math"or a considered judgment that little would be gained by waiting for more tests of low template DNA (as the court-appointed experts advised him)? That was the question addressed in discussions here back in March and May.

Now, an article in the Proceedings of the National Academy of Sciences of the USA reports that a different mixed profile does not seem to reflect Amanda Knox's DNA. Seeking to clarify a New Scientist news report entitled "Software says Amanda Knox's DNA Wasn't at Crime Scene," Professor David Balding, a statistical geneticist at University College London who has made important contributions to the statistical evaluation of forensic DNA profiles, describes this finding on his website as follows:

I do reanalyse a crime scene profile from the murder of Meredith Kercher in Perugia, Italy, in 2009 and find no support for the presence of DNA from Amanda Knox as they say, but the main question for that item was whether there is DNA from Raffaele Sollecito, and my analysis finds strong support for this even allowing for many of the problems with the DNA evidence highlighted at the appeal in 2011.

As the PNAS article makes clear, Dr. Balding used software that he is developing for LT-DNA profiles to evaluate the "five loci from exhibit 165B of the trial in Perugia, Italy, in 2009. The exhibit includes the clasp of a bra, attached to some apparently blood-stained fabric, that was found near the murdered woman, Meredith Kercher."

The prosecution's expert testified at the trial that the profile of the clasp DNA matched all the alleles of the victim and one of Sollecito's. The court's experts (Vecchiotti and Conti) agreed "but also reported many additional epg [electropherogram] peaks. ... Of the 24 additional peaks ... , of which 6 had heights below the threshold of 50 relative fluorescence units, 9 are included in [Knox's] profile ... providing apparent support for the presence of DNA from her. However, four of her alleles were not observed, including two homozygotes, which are less prone to dropout [in such small samples]." In other words, it is hard to know what to make of the data by eyeballing it.

The statistical software (to the extent I understand it) models allelic "drop in" and "drop out" and uses allele frequencies in the population to generate likelihood ratios for the peaks that the analyst regards as the result of true alleles as well as those designated as "uncertain" alleles. In this case, the well established alleles indicated a mixture of DNA from three or more people, and Balding considered two competing hypotheses: (1) the DNA is a mixture from Kercher, Sollecito, and one unknown individual vs. (2) the DNA is from Kercher and two unknown individuals.

He found that the perceived pattern of alleles was 42 million times more probable if Sollecito's DNA was there. As for Knox, however, "separate analysis with her as the queried contributor returned an LR < 1." That is, the designated pattern would arise more often if Knox's DNA were in the mixture than if it were not -- a finding "favoring a conclusion of no DNA from her."

What this means for the case depends on what some forensic scientists call "activity level" hypotheses and what lawyers call "relevance." What activity would cause Sollecito's (but not Knox's) DNA to be detected on the bra clasp (but not on the fabric)? The defense theory, of course, is that neither of them touched the bra, but Sollecito's DNA molecules were transferred there at a later time. The prosecution theory could be that Sollecito (but not Knox) touched the bra during the murder.

Related postings

• Are Two Heads Better than One? Mathematics and the Amanda Knox Case, Mar. 27, 2013
• Bad Math or Passable Law? DNA Testing in the Continuing Prosecution of Amanda Knox and Raffaele Sollecito, May 25, 2013