Nonindependence of matches at different loci in DNA profiles: quantifying the effect of close relatives on the match probability.

In comparing a particular DNA profile with that from an unknown (but distinct) individual, matches at different loci between the profiles will not be independent, even in a randomly mating population, because of the presence in the population of relatives of the individuals. The paper contains a theoretical analysis of the extent of this effect on the match probability, for profiling techniques which separately probe different loci. Naive calculation using the product rule could substantially understate the match probability. Past a certain point, the testing of additional loci provides no more information than would be available in discriminating between sibs. The correlation effect described here would be unimportant in criminal casework if close relatives of the suspect, and in particular full-sibs, were excluded as possible culprits. In the absence of such exclusions the current practice of effectively ignoring such relatives in presenting match probabilities could be extremely prejudicial to a suspect, even in cases in which there is no direct evidence to incriminate his/her relatives.