Knox/Sollecito: bra-clasp DNA printout

Knox/Sollecito bra-clasp printout version 1

Click on the image for a larger version

If you click on the image to the left you’ll see a slightly annotated version of the computer printout for the DNA on the bra-clasp, taken from the review of DNA evidence in the appeal of Amanda Knox and Raffaele Sollecito.

Firstly, I’ve added coloured dots to show the position of alleles relating to Sollecito and to Meredith Kercher.

It should be noted, though, that, while Kercher’s full DNA profile is known, Sollecito’s is not. It is possible to partly work it out, because the review contains details relating to the mixed Kercher/Sollecito sample. According to the Massei report, this sample yielded a match to Sollecito across all 16 of the loci (groups of alleles) tested. Given this knowledge, we can subtract Kercher’s profile from the profile of the mixture, which leaves us with alleles relating to Sollecito.

However, some of Sollecito’s alleles overlap with Kercher’s. I’ve marked these on the image with pale blue dots. It’s unfortunately not possible to be sure which of these are genuine Sollecito alleles, but I’ve also added orange dots for alleles which are common in south Asia and rare in white Europeans (because Kercher’s mother is South Asian, it is likely that these are Kercher’s alleles but not Sollecito’s).

Sorry if anyone’s disappointed not to see a definite Sollecito profile there, but you can console yourself with the knowledge that the probably only reason this would be useful would be for checking the DNA profile on the bra-clasp. And, since using a DNA profile based on the bra-clasp testing in order to check the bra-clasp testing would be a bit circular, you’re not really missing out on much.

Marked with a figure 1 on the image are the peaks which Conti and Vecchiotto, authors of the DNA review, suggest ought to not to be rejected as “stutters” (a type of noise) but instead considered as potentially genuine alleles. I’ll examine this issue further in another post. But, to cut a long story short, Conti and Vecchiotto follow a recommendation in this report that peaks should not be rejected as stutters if they are both over 50 RFU in height and also over 15% of the height of the next known allele to their right.

Hope you like my dots. Just to warn you: for my next trick I propose to add green dots representing the DNA profile of Amanda Knox.


14 Responses to Knox/Sollecito: bra-clasp DNA printout

  1. Hi Maundy,

    I would not publish Mr. Sollecito’s or Ms. Knox’s full reference profile as a matter of privacy.

  2. Hi Maundy,

    It is contrary to practice among forensic geneticists to have the reference profiles while performing the analysis. The process of “sequential unmasking” of the reference information is one alternative when analyzing mixtures, as proposed by Dan Krane and coauthors.

    • maundy says:

      Hi Chris.

      I’d take the stance that I’m not actually performing any DNA analysis here. I’m just putting dots on the printout for easy reference.

      What I’m not doing is attempting to make decisions about what peaks to include or not include, which would indeed be bad practice with the reference profiles to hand.

      I’m also not an actual DNA scientist, you know!

  3. Julia says:

    You say the Massai report says it located 16 loci. I compared Amanda and Meredith’s DNA across 30 peaks and they share 9 peaks. Meredith’s DNA on the knife is 30 peaks. So they usually test 30 peaks, correct?

    Does the Massai report say that the 16 peaks located were unique to Sollecito and not Meredith? Because I think they share 11 peaks.

    If you are right, as you stated previously, that 16 peaks means it’s 100% you no matter what other peaks are missing, then I don’t understand why they normally test 30. It seems like 14 would be too many to be missing.

    I think in order for you to really make your argument work, you need to find a random person’s profile and show how their profile is not on the bra clasp.

    • maundy says:

      Hi Julia,

      I’m not sure how much that is a comment on the post above and how much on the idea of carrying out a comparison to Knox.

      You’re right that Knox and Kercher share 9 peaks in their profile. That’s something that has to be taken into account when working our a random match probability. You can’t just work out the chances of a particular person’s DNA being there – you have to do it in the context of the mixture.

      Loci and peaks are different things. In a full reading, each locus has usually two peaks, sometimes one peak, relating to an individual whose DNA is present. But scientists seem to usually deal in terms of matching loci, rather than individual peaks (ie if you were able to match one peak at each locus, that would not normally be considered “half a match”).

      Your idea of also comparing a random profile is a good one. I’ll think about whether I can do that. Thanks for the idea.

  4. Actually it was the very same issue. The scientist with whom I consulted wrote, “It only takes a single allele difference to constitute an exclusion.”

    The peaks on the bra clasp can be assigned to individual contributors in more than one way, making the story more complex.

    • maundy says:

      I think you should check back with this scientist, Chris. In a sample from a single subject/known mixture, one allele difference should indeed result in a no-match. But the thing to understand is that, once you have a large number of loci that match, suddenly finding one that doesn’t is statistically impossible. Or, it means that your sample must not be from a single subject/known mixture after all.

      • Hi Maundy,

        It might help to define more precisely what you mean by single subject/known mixture. Find me a single forensic geneticist that agrees with your statement, and I will ask the forensic geneticists that I know to comment on it. One problem with a statistically-based argument such as yours is that it can become a license to bend the data to fit the interpretation, in other words bias can creep into the interpretation.

  5. cjp says:

    Excellent idea, Julia. I admit that I am one of those laypeople who’s experience with DNA to this point has been “CSI”.
    I’ve lived a long time, and, until this case, I had no idea exactly what went into discovering “the DNA is a match”.
    I’m not ashamed to admit that my eyes glaze over when trying to get my head around some of these explanations.
    I think a random sample would go a long way in showing how a match is determined, I hope you can do this, Maundy. Thank you

    • maundy says:

      I guess it might be a glaze-a-thon for quite a few people. After my next post, I’ll try to steer back to not talking about DNA profiles.

      You can see the random match in the next post after this one.

  6. Hi Maundy

    John Butler (p. 386 in Forensic DNA Typing) wrote, “In forensic DNA typing, if any one STR locus fails to match when comparing the genotypes between two or more samples, then the profiles between the questioned and reference sample will be declared a non-match, regardless of how many other loci match.” Paternity testing is an exception, but it is a comparison between two individuals, and a mutation may have occurred.

    • maundy says:

      Chris, I think you’re missing the whole point. Yes, it’s theoretically correct, in the same way that if you were to see a person who looked exactly like me but had differently coloured eyes then it wouldn’t be me. But the more important point is that we can be sure that no such person exists, so the situation is not going to arise. Similarly, no two people will ever be found whose DNA samples match on 15 of 16 loci but not on the 16th.

      In the present case, we don’t have a mismatch at any locus.

  7. Hi Maundy,

    With Meredith and Raffalele sharing 11 alleles, one could certainly obtain a partial profile of Raffaele. To obtain a full profile of 15 loci one would have to have a means of deconvoluting each person’s contribution to the intensities of the peaks at those 11 alleles. Count me as one who is skeptical that this can be done in a completely objective and relatively trustworthy manner for reasons that I have already given.

    As to the question of having a match at every locus, that assertion was disputed by Dr. Tagliabracci at the trial of first instance. I am not saying that I agree with his analysis, only that the issue is not as cut-and-dried as you imply.

    Tangential the subject of the present case but quite pertinent to DNA analysis as a whole is the mathematics of the birthday problem, addressed at least twice by mathematician Keith Devlin. For a database of a million people and a 9-locus profile, the odds are better than 50% that two people will have identical profiles. Some databases are larger than this already, and they are growing. Devlin wrote, “There may be no way of avoiding bringing mathematicians into court to explain how the calculations are done. But for that to be effective, those judges and juries need first to learn (and accept) that human intuitions about probabilities are hopelessly unreliable.”

    • maundy says:

      Hi Chris. Your concerns about overlapping alleles seem to me to amount to scepticism that it is possible to identify the contributors to a mixed sample. See p.158 of Butler for material relating to that. For my calculation here, I allocated the overlapping alleles a probability of one (i.e. we know from the outset that they will be found to match).

      Tagliabracci’s approach was different from C&V. They question only 4 loci and they do not dispute that there is a full match to Sollecito. Their hypothesis is, instead, that there may be additional DNA. Tagliabracci’s approach was really to try to cast doubt on as many loci as he could get away with. If you can get the number of matching loci down into single figures then you maybe start to get into the realms of suggesting misidentification (even the odds will still be massively against it in reality).

      Birthday problem: This works on the basis that, in a given population you are looking for a match between any two people. In a population of a million, you can do this using any combination of two people, so you have a trillion chances to make your match. In forensic DNA, though, you are trying to find a match to a specific reference profile and, when you have an actual suspect (as opposed to trawling through a database) you only have one chance to make the match. So, the birthday problem doesn’t have any effect on the odds in this specific real-life example.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: