Was Casey Anthony’s DNA on the bra-clasp?

DNA readout from bra-clasp in Kercher case

Click on the image for a larger version

In 2009, the journalist Baribie Nadeau mentioned something in an article which many people will have wondered about when they read it. She said that Vincenzo Pascali, a scientific consultant for the prosecution in the early days of the process, had “hinted” that a sample taken from the bra-clasp of the murder victim Meredith Kercher had contained the DNA of not only Raffaele Sollecito but also his alleged accomplice in the crime, Amanda Knox.

It may be thought unlikely that there was much in this. That’s not to impugn Nadeau’s journalism – if a DNA expert hints at something like that then it is worth reporting. At the same time, though, you would think that, if Knox’s DNA really were present on the clasp, the prosecution would have mentioned it by now.

Nevertheless, given the information that is now available about DNA evidence in the case, I thought it would be worth looking into this suggestion.

The first thing to note – and it’s an extremely important thing to note – is that the DNA of Kercher and Sollecito found in the sample from the clasp accounts for either all or very nearly all (depending on which expert you are listening to) of the alleles observed. Since there is so little left to play with, trying to make a further DNA match out of it would seem at bit like looking at a nine and a six and wondering whose phone number in particular they might form part of. Completely impossible, even if you believe that further DNA might be present. Impossible, that is, using any method that is generally considered scientifically reliable.

What you could do, though, is lower your standards and begin to look at peaks that would ordinarily be ignored on the grounds that they could well be nothing more than noise. This makes the peaks that you are including less reliable (how much less reliable depends on how radical you are in including smaller peaks), but it might be argued that there is a legitimate trade-off. If you can get a very good match, then maybe it can be argued that a questionable peak here or there is acceptable. After all, how likely is it that the noise has simply been kind enough to appear in the exact places you would need it to in order to make you incomplete match complete?

So, you can see on the image that I’ve added green dots to the DNA profile taken from the bra-clasp, indicating the places where Knox’s alleles would be expected to be located. And it turns out that it is possible to see a fairly good degree of compatibility with the sample. But I had to go all the way and include even the tiniest bumps on the printout. Because some of these are so small that no numerical information about them is included, I also had to judge a few of the locations by eye (the blue numbers on the image indicate where I did this). This raises the additional problem that my own human error may have some effect on the reliability of what I have done.

But how strong is the actual match? Well, first you have to explain away two things. At loci D8S1179 and D19S433 (the first and tenth groups of peaks on the image), there are small peaks which the DNA review says are above the threshold where we can be confident they are noise. But they are only above the threshold by a hair’s breadth. So maybe we can consider the possibility that they really are noise. Then at locus D7S820 (the third group of peaks in the image), there’s location where you would expect to see an allele relating to Knox, but there is nothing there. Again, maybe this is not so much of a problem. We’re presumably dealing with an extremely faint DNA trace. Maybe it’s simply the case that just one peak out of 29 is so faint as to not be there at all.

What’s starting to emerge, though, is the problem with not having any real standard to go by in terms of which peaks you consider to be real and which you don’t. The judgements you have to make in order to determine whether there is compatibility or not start to become fairly subjective, at least in this case. At each locus point, the ways in which compatibility could arguably be ascertained are numerous, so things are different from how they were in considering the matches to Sollecito and Kercher. Whereas their profiles are quite clear and match 100% to unquestionably genuine alleles on the graph, it seems like you can pretty much see Knox’s DNA profile in the printout according to how much you want to.

A commenter suggested adding dots for a sample known not to be connected to the case. I thought this was a good idea, and whose DNA profile could be more suitable for the task than that of Casey Anthony, recently acquitted of murder in the US? So I’ve added her alleles using pink dots.

I think this goes to further illustrate the problem. Although there are a greater number of problematic spots on the graph that you need to ignore if you want to make Anthony’s DNA fit, some things go the other way. For example, 69% of Knox’s allele’s pair up with peaks on the graph that are above 50 RFU in height (one measure of a potentially genuine allele). It may sound from that like Knox’s DNA is beating the odds. But Anthony’s DNA profile, on the other hand, matches to those >50 RFU peaks 75% of the time. For me, it seems hard to see how you could justify being confident that Knox’s profile is there without also thinking that there is a least a decent chance that Anthony’s could also be there.

Then again, it’s always possible that I have it backwards and that, in reality, the bra-clasp printout offers a new chance for that half of America outraged by Anthony’s aquittal to get her back in court.


7 Responses to Was Casey Anthony’s DNA on the bra-clasp?

  1. cjp says:

    Great post. What this layperson gets from this explanation is:
    While it would be problematic for the prosecution to insist that the DNA on the bra clasp belongs to Amanda, there isn’t anything here that would give the defense an AHA! moment and say, It can’t possibly belong to Amanda”. This evidence does nothing to exclude her. Do I have this right?

    • maundy says:

      Yes, sort of. The defence are not able to say it definitely doesn’t belong to Knox. However, I’d put it more strongly than “problematic” in terms of the prosecution saying it does belong to Knox. It looks to me like there is no real scientific basis for making such a claim.

      What should be noted is that the prosecution have never actually said that. So there’s not really any big revelation in what I have done.

      By contrast though, if you compare the Casey Anthony dots to the Sollecito dots, you can see a big difference, with Sollectio showing 100% perfect compatibility.

  2. Julia says:

    Am I right in my glance at this that Casey Anthony matches 13 loci? And of the remaining 3 she has “bumps” on 2 and is absent from the last? Also, she registers significant alleles on 24 out of 30?

    Sollecito MIGHT match 11 alleles of Meredith’s. I’m not certain since his actual profile hasn’t been published. Did his remaining 19 allleles match? Does he match 16 loci? If, hypothetically, only 24 out of 30 of his allele’s can be certain, and 24 out of 30 are certain with Casey Anthony, then it’s just as likely to be Casey Anthony’s DNA as it is his?

    I was reading that another problem is that certain alleles of his are in places where Meredith’s DNA would cause stutter.

    I was looking this same data over separetely, and it APPEARED to me that all 19 of Sollecito’s unique alleles were present, but I can’t read italian and it became too difficult to ascertain whether the mixed profile included only his DNA and Meredith’s or what.

    Thank you so much for these very interesting blogs. Including Casey Anthony’s DNA makes this much easier to follow and compare. I wouldn’t have thought she would match up 75%.

    • maundy says:

      Hi Julia,

      In terms of the numbers, Anthony matches 10 loci out of the 16. You can’t count loci where there is a figure 2 (no peak) or a double black dot (just a blip). For seven of the ten, you have to consider that alleles relating to either Kercher or Sollecito are already present at all points where you would expect to find matches to Anthony, so those seven don’t really tell us very much (they are “matches” to Anthony before you even begin). At the remaining three, there is always at least one allele relating to Kercher or Sollecito already present. Overall, there are only two peaks which are large enough to be reliable and can be associated only with Anthony, plus an additional one peak which both Stefanoni and the DNA review consider to be a stutter.

      In Sollecito’s case, he matches at 16 of the 16 loci, with no figure 2s or black dots. Probably 10 (IMO) of his peaks are overlaps with Kercher and the remaining 17 peaks are attributable to Sollecito alone. The overlaps have the effect of increasing the chances of a random match (i.e. the chances that the trace is not Sollecito). I took the overlapping peaks into account in the calculation I did in this post, and still came up with a ratio of 22 billion to one.

      In terms of whether any of Sollecito’s peaks might actually be stutters: As far as I understand, any peak which has a genuine allele reasonably close to its right hand side is in a position where it might potentially be a stutter. You can see from the image that this applies to most of Sollecito’s peaks. However, according to the DNA review, a peak is not a stutter if it is more than 15% the height of the allele to its right. Assuming that is correct, none of Sollecito’s peaks are stutters.

  3. Julia says:

    Yes, but your probability is based on the concept that there are only 2 profiles on the clasp. Or is it not? If there can only be two peaks per loci per DNA profile, and some of these loci show more than 4 (fairly clear that it is more than 4), then the ratio is changed. Finding 17 peaks out of 19 to 20 possible unique peaks seems like a slam dunk to me, I’d agree. However, there shouldn’t be 19 possible unique peaks to Sollecito. There should be less than that since there is at least one other profile on the clasp.

    Say that the two profiles were Casey and Amanda. That has left 11 unique peaks for Sollecito. Casey overlaps his profile in 6 places and Amanda in 4 places, am I correct? (My eyes are blurring with all these dots so I need fact-checking).

    It still seems to me that his profile must be on there, but aren’t we missing a few peaks? If you have 27 peaks listed there, what about the other 3? (Or are the other 3 the contested ones?)

    I wonder at which point the probabilities don’t work to identify anyone. i.e. if you have 6 profiles it is too crowded to figure anything out.

    • maundy says:

      Hi Julia,

      I excluded the loci with more than 4 possible peaks from the calculation. I don’t just throw this stuff together, you know!

      Talking about Sollecito having “overlaps” with Anthony is problematic, because Anthony’s DNA is not actually there. We can talk about overlaps with Kercher because her DNA is very clear on the printout (it’s all the really tall spikes) – that means the relevant peaks relating to her DNA are already present. But there’s no reason why having an overlap with Anthony has any greater an effect in terms of probabilities than having an overlap with, say, Morgan Freeman or Justin Beiber. The difference being that no peaks from them can be said to be already present.

      Does that make sense?

      On number of peaks: there’s no set number of peaks for someone to have over the 16 loci. Sometimes there’s two peaks at a loci, sometimes one.

      I think the chances of a match increase each time you add someone. I think it’s not normal to go above a three-person mix, though maybe its possible. The real problem with trying to identify Knox or Anthony in the image is not really that, though. It’s the fact that you end up trying to make use of tiny peaks which you should really be ignoring.

  4. Maundy, This exercise is a great illustration of why forensic geneticists should never analyze a piece of evidence with foreknowledge of the reference profile of suspects.

Leave a Reply

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

WordPress.com Logo

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

Google photo

You are commenting using your Google 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 )

Connecting to %s

%d bloggers like this: