Likelihood ratios for evaluating DNA matches obtained from a database search when there is substructure in the population.

DNA database searches are frequently conducted to identify the suspects of crimes. When a match is obtained from such a database search, the evidence must be evaluated. Existing methods for assessment of the DNA evidence in this scenario require assumption of stochastic independence between the profiles in the database. However, when there is substructure in the population, this assumption is violated. The problem of how to account for population substructure in the database search scenario is analyzed and a solution in the form of a bounded estimate for the likelihood ratio is presented. The implications of these methods are investigated in a realistic scenario using published forensic allele frequencies to simulate ten-locus DNA profiles. In the simulated example, it is observed that the strength of the evidence can be inflated by more than a factor of 10 in 11.6% of database search cases if mild population substructure is ignored. With these methods, the magnitude of the subpopulation effect in the database search scenario can be quantified and the weight of the evidence of a DNA match more accurately assessed.