Summarizing DNA evidence when relatives are possible suspects

Summaries of deoxyribonucleic acid (DNA) evidence in criminal proceedings have been controversial in part because calculations of the probability of a ''match'' between crime and suspect samples assuming that the pairing of these samples occurred by chance typically have depended on independence or conditional independence assumptions that are subject to dispute. This article describes a new methodology that summarizes DNA evidence by addressing the possibility that a relative of the accused individual is the source of a crime sample, which can be viewed as complementary to assessing the probability of a ''match'' assuming that individuals in a broader population are paired by chance. The new technique uses a statistical criterion based on distances between bands from autoradiograph images to distinguish individuals without relying on any conditional independence assumptions. We apply the method to data from a paternity testing laboratory to illustrate its ability to distinguish between the DNA profiles of related individuals. In addition, a simulation study suggests that to distinguish reliably between family members, more genetic loci should be probed than the three to six loci typically investigated in current practice.