Amanda Knox and Raffaele Sollecito and the murder of Meredith Kercher, Part I

Introduction to the case

This post will be the first in a series of articles on the Amanda Knox/Raffaele Sollecito case. Amanda Knox is an American student, and Italian Raffaele Sollecito is her former boyfriend. Both of these two individuals were convicted of the murder of British student Meredith Kercher in late 2009. Rudy Guede was previously convicted of her murder in a separate, fast-track trial in 2008. Many in the United States have criticized the prosecutor’s summation (http://blog.simplejustice.us/2009/12/05/a-trial-without-evidence.aspx) and the forensics (http://www.sciencespheres.com/2009/10/methods-of-polizia-pseudoscientificaa.html).

The DNA forensics in this case centers around two items, Meredith Kercher’s bra clasp and Raffale Sollecito’s kitchen knife. This article will focus on some of the issues surrounding the knife. The kitchen knife had Amanda’s DNA on the handle, which is not surprising given that Amanda cooked at Raffaele’s house. But the prosecution claims that their forensics investigators found a small amount of Meredith’s DNA on the blade. This claim is dubious at best, but even if it were true, the knife cannot be the sole murder weapon. Let’s take up the second question first.

Problems with the knife

First, Mr. Sollecito’s kitchen knife was too large to have made two of the three wounds on Ms. Kercher’s body. The smaller knife, the one that made the first two wounds, may have also made the third wound. Second, the kitchen knife does not match the bloody outline of a knife at the crime scene (http://www.newsweek.com/id/216903/page/2). Third, Newsweek reported that, “an officer testifying at the trial said he used ‘police intuition’ when choosing that knife from Sollecito's cutlery.” If there were multiple knives in the drawer, why choose one that had a different outline from the one whose outline was on Ms. Kercher’s sheets? The claim of police intuition does not make any sense.

But possibly the most serious reason for doubting that DNA was really on the knife is that it tested negative for blood (http://www.sciencespheres.com/2009/12/why-knife-was-not-knife.html. Dr. Stefanoni opined that the knife had been cleaned with bleach. This is a puzzling claim even at the outset, because bleach does not leave a corrosive mark on stainless steel; however, the prosecution’s argument becomes even more questionable upon further inspection. Drs. Elizabeth A. Johnson and Greg Hampikian, both experts in DNA forensics, coauthored an open letter about some of the evidence in this case (http://www.friendsofamanda.org/articles.html). About the cleaning of the knife, they said:

This DNA does not originate from blood. A highly sensitive chemical test for blood was negative, and it is unlikely that all chemically detectable traces of blood could be removed while retaining sufficient cells to produce a DNA profile consistent with the victim.

Indeed, bleach is so effective at destroying DNA that it is used in research laboratories for that very purpose (A. M. Prince, L. Andrus PCR: How to kill unwanted DNA, Biotechniques, Vol. 12, No. 3, 358-360). It far more effective than acid! Bleach is also used in some anthropological work to destroy unwanted DNA on the surface of an object that would contaminate valuable DNA inside. The very fact that there was no blood on the knife suggests that there was also no DNA on the knife; therefore, contaminating DNA is the most likely source for the signals that were observed. These three reasons call into question whether the kitchen knife had anything to do with the murder.

What the laboratory of Dr. Patrizia Stefanoni should have done is to save some of the material for a retest. As mentioned previously, this lab should have tested other knives, and perhaps other, random implements, in Raffele Sollecito’s drawer in the same way. Would DNA have also shown up on a different knife? These control experiments would have clarified whether Meredith Kercher’s DNA was really on the knife or not. However, we can at least examine the data that were produced from the knife.

Introduction to DNA forensics

Modern DNA forensic analysis (birg.cs.wright.edu/talks/Human Identification 2008.ppt) produces fluorescent signals that must be observed and interpreted properly to exclude or not exclude a person as a DNA contributor. The fluorescent signals arise from a collection of DNA molecules of various lengths (sizes) that are produced by chopping the DNA strands at a set of specific locations, and then separated using capillary electrophoresis. The pattern of signals from the DNA form what is called an electropherogram. The electropherogram from a piece of evidence is compared against reference samples from various individuals.

Fluorescence Spectroscopy

Spectroscopy is the study of how light interacts with matter. All spectroscopic experiments involve the observation of photons of light above a background of random noise. All forms of spectroscopy must contend with the fact that signal-to-noise (S/N) ratios are not infinite. The S/N ratio is a measure of the strength of the signal and is often related to the concentration of the chemical species that produces it. DNA forensics makes use of chemical labels that produce fluorescence. In a DNA electropherogram the S/N ratio is measured in relative fluorescence units (RFUs). Why are S/N ratios so important in regards to the DNA sample on the knife blade in the Amanda Knox/Raffaele Sollicito case?

The open letter coauthored by Dr. Elizabeth and Professor Gregory Hampikian and cosigned by several others had this to say about the knife DNA:

An extremely low level, partial DNA profile was developed for the blade swabbing using the Identifiler kit. The alleles detected were consistent with the DNA of the victim. The highest peak in the electropherogram was approximately 100 relative fluorescence units (rfu), while 21 of the 29 peaks that were detected and labeled as alleles fell between 20 and 50 rfu….No credible scientific evidence has been presented to associate this kitchen knife with the murder of Meredith Kercher. (emphasis added)

Signal-to-Noise Ratios

To understand their conclusion we need to know more about signals and noise. Suppose you are driving your car away from your home town and you have tuned in your favorite FM radio station. As you travel farther away, the music gets fainter (the signal is now weaker), so you turn up the gain (amplification). Now the music is more audible, but you start to hear crackles (the noise). Amplification affects both signal and noise. Eventually, you will travel so far away that the noise becomes more unpleasant and you switch to a different station (the S/N ratio has become unacceptably low).

Now let us ask what would happen if you were 200 miles away and you tuned to the frequency of your favorite FM station in your home city. You would hear nothing but noise, sometimes called static. Based only on your observation, you could not say that the station was even broadcasting. It is not that you would deny that the station was broadcasting, it is just that you cannot affirm it on the basis of your observation; your observation is indeterminate.

Each DNA forensic laboratory establishes a threshold value for the size of acceptable peaks, but the threshold values are not identical from lab-to-lab. Setting a minimum threshold does not automatically favor the defense or the prosecution. The lowest such value of which I am aware is 40 (http://www.bioforensics.com/articles/champion1/champion1.html). At first this seems large, since peaks smaller than this are still greater than noise, but these peaks are small relative to those typically encountered in DNA electropherograms.

Instead of examining the specific choice of a threshold value, let us discuss why setting them in advance of the experiment is so important. Scientists set up their experiments to test (falsify) their hypotheses. If the signal-to-noise level in any spectroscopic experiment falls below the threshold, any hypothesis requiring that the signal be above the threshold must be rejected. The reason one sets the threshold first is to avoid bias. The textbook An Introduction to Forensic DNA analysis, 2nd ed. (Rudin, N. and Inman, K., CRC Press 2002, p. 121) states (emphasis added), “It is important to have some predetermined limit to distinguish what is signal and what is noise.” If one treats the threshold as flexible, the very purpose for setting it prior to running the experiment is defeated.

How does this relate to the DNA electropherogram of the knife sample (http://www.friendsofamanda.org/articles.html, Figure 1)? 16 out of 29 peaks are lower than 40 in S/N, and 22 are lower than 50 RFU. Dr. Stefanoni herself wrote (http://www.thedailybeast.com/blogs-and-stories/2009-09-14/no-mistrial-for-knox/) the words “too low” with respect to the knife. In other words these peaks are too small to count as music in the car radio analogy above.

It would be helpful to know the threshold value regularly employed in Dr. Stefanoni’s lab, but it could have been as high as 150 RFU and is very unlikely to be below 40 RFU. The signals above the threshold might constitute a partial profile of Ms. Kercher’s DNA. However, such a partial match is very problematic (http://www.scientific.org/tutorials/articles/riley/riley.html):

1. A partial profile essentially proves that one is operating outside of well-characterized and recommended limits.
2. Contaminating DNA usually presents as a partial profile, although not always. For this reason, the risk that the result is a contaminant is greater than for samples that present as full profiles.
3. A partial profile is at risk of being incomplete and misleading. The partial nature of it proves that DNA molecules have been missed. There is no way of firmly determining what the complete profile would have been, except by seeking other samples that may present a full profile.

Conclusions

The kitchen knife cannot be the sole murder weapon, and it most likely had nothing to do with the crime whatsoever. The only evidence that it might have played any role is the finding of DNA that is at best a partial match to Ms. Kercher’s DNA on the blade, and that almost certainly came from contamination during the test. Given the weakness of the fluorescence signals and especially the lack of blood on the knife, the signals observed do not tie the knife to the crime. No wonder that the open letter quoted above ends its section on the kitchen knife by concluding that there is no forensic evidence to link it to the crime. But an equally troubling problem with the prosecution’s theory is that Ms. Knox and Mr. Sollecito would bother to take this knife from his flat to the dwelling occupied by Ms. Knox, Ms. Kercher, and their two roommates in preference to using knives found in the kitchen of these four roommates. What sense does that make? It is typical of this case that Occam’s razor is so often ignored.

Update, 01/18/2010

The prosecuation's theory would also have us believe that Knox would not dispose of the supposed murder weapon, and despite cleaning the blade extensively, would not clean her fingerprints from the handle. This strikes me as unlikely.