An article posted yesterday at Technology Review talks about some recent research which, its authors hope, will lead to the development of a new type of forensic tool. The research (also reported by Ars Technica and Science Daily) was carried out at the University of Colorado at Boulder, and looked at the feasibility of using bacterial samples gathered from ordinary objects (in this case, computer keyboards and mice) to identify the person who had used them. The idea is based on the observation, which I’ve mentioned before, that each of us carries around a large number of bacteria and other microorganisms, essentially as part of us. The exact composition of this menagerie varies from person to person, so the idea here was that the genetic profile of the microbes on a person’s skin might be a useful identification tool.
The results of the experiment did give some support to the idea, although further investigation is clearly required. The sample sizes used were quite small — the larger of the two trials used samples from only nine computer mice, compared to a database of 270 individuals. That it was possible to associate the mice with their users is promising, but there are many questions that remain to be answered.
First, it is not known at this point how stable a person’s “microbial fingerprint” is over time. It is known that use of antibiotics, and even hand washing, can disrupt the profile temporarily; how long the disruption lasts is also an open question. There are also questions about the stability of the traces left behind on objects. The research team did some work that suggested that the traces were somewhat persistent; however, it is not clear how use of the object by another person, or even routine cleaning, would affect this. (Considering the lack of hygiene I have seen in some university computer labs, one might end up concluding the last user was Typhoid Mary.)
If this is considered as a potential forensic technique, there is a much more serious information deficiency. As far as I know, no one has any reliable data on the degree to which microbial profiles vary across the population, so there is really no way to evaluate the probability of a match purely by chance. (I’ve mentioned before that, even with DNA evidence, the data is less comprehensive than many experts would like.) Getting only one match in a sample of 270 does not provide very strong evidence; extrapolation of a probability distribution from one data point is generally frowned upon.
We may see this appear on television programs, like the CSI series. (These things always work better on TV; facial recognition software works all the time there, even though its real-life performance is mostly unblemished by success.) But I think (and hope, actually) that it will be a while before this becomes an accepted forensic technique. It does, though, give you another reason to wash your hands.