DNA profile match probabilities in a subdivided population: when can subdivision be ignored?

Li and Chakravarti [Li, C.C. & Chakravarti, A. (1994) Hum. Hered. 44, 100-109] compared the probability (MO) of a random match between the two DNA profiles of a pair of individuals drawn from a random-mating population to the probability (MF) of the match between a pair of random individuals drawn from a subdivided population. The level of heterogeneity in this subdivided population is measured by the parameter F, where there is no subdivision when F = 0 and increasing values of F indicate increasing subdivisions. Li and Chakravarti concluded that it is conservative to use the match probability MO, which is derived under the assumption that the two individuals are drawn from a homogeneous random-mating population without subdivision. However, MO may not be always greater than MF, even for biologically reasonable values of F. We explore here those mathematical conditions under which MO is less than MF, and we find that MO is not conservative mainly when there is an allele with a much higher frequency than all the other alleles. When empirical data for both variable number of tandem repeat (VNTR) and short tandem repeat (STR) systems are evaluated, we find that in the majority of cases MO represents a conservative probability of a match, and so the subdivision of human populations may usually be ignored for a random match, although not, of course, for relatives. Loci for which MO is not conservative should be avoided for forensic inference.