A postdoc in the group, Rori Rohlfs, has published a paper in PLoS Genetics on familial identification in forensic genetics. When a DNA sample has been obtained from a crime scene, it is common practice in the US to search databases of DNA profiles for possible matches. Recently, this practice has been extended to familial searching. Forensic investigators will search databases, not only for complete matches, but also for individuals with partial matches who may be related to the person who left the sample at the crime scene. Rori Rohlfs and her collaborators (including her former advisor Bruce Weir) provide an analysis of the statistical aspects of this approach. They show that individuals for whom the reference databases are less representative, or individuals originating from population groups with reduced genetic diversity or over-represented in the offender/arrestee DNA profile databases, are disproportionately more often subject to false familial identification. In other words, individuals from smaller groups, or groups that are over-represented among arrestees, are more likely to be falsely investigated for crimes they have not committed. You can find a blog entry with an animated cartoon explanation of the problem here, a news story about it here, and the original scientific publication is here.
While you could argue that the police should use any means available to them in the investigation of serious crimes, even if it occasionally will lead them down the wrong path, the use of forensic methodologies that create biases against particular population groups is, needless to say, very worrying. The results by Rori Rohlfs and her collaborators put a serious question mark to the practice of familial identification. They conclude in their scientific paper that “care is warranted in the use and interpretation of familial searching forensic techniques”. That is a scientists way of saying that this is seriously messed up and something should be done about it.