Allele frequencies and minor contributor match statistic convergence using simulated population replicates

Probabilistic genotyping permits a comparison of forensic evidence given hypotheses regarding the origin of observed short tandem repeat alleles in a mixed DNA profile. Using the publicly available R package forensim, it has been proposed that mixtures with non-contributors from low genetic diversity populations are more likely to be mistakenly identified as contributors to a mixture than non-contributors from high genetic diversity populations. We hypothesized that these observations are attributed to the unique distribution of alleles in the reference population and may not generalize to other samplings of the same population. We used forensim to simulate 200 US populations (50 each of self-reported African-American, Asian-American, European-American, and Hispanic descent). We compared likelihood ratios for 2400 mixtures to those derived from published data and identified stark differences. A minimum of ten population replicates were required to reduce observed differences relative to published data. Deviations from Hardy–Weinberg equilibrium and allele frequency distributions suggest that simulated populations should be sufficiently evaluated for expectations of population genetic parameters prior to use in DNA mixture modeling experiments. Overall, our findings support the utility of forensim and further describe its suitability to model population genetic parameters but suggest that a single population replicate (directly ascertained or simulated) may be insufficient to make conclusions about a given DNA mixture.