Geolocation prediction from STR genotyping: a pilot study in five geographically distinct global populations.

Background: Traditional CE-based STR profiles are highly useful for the purpose of individualisation. However, they do not give any additional information without the presence of the reference sample for comparison.

Aim: To assess the usability of STR-based genotypes for the prediction of an individual's geolocation.

Subjects and methods: Genotype data from five geographically distinct populations, i.e. Caucasian, Hispanic, Asian, Estonian, and Bahrainian, were collected from the published literature.

Results: A significant difference (p < 0.05) in the observed genotypes was found between these populations. D1S1656 and SE33 showed substantial differences in their genotype frequencies across the tested populations. SE33, D12S391, D21S11, D19S433, D18S51, and D1S1656 were found to have the highest occurrence of "unique genotype's" in different populations. In addition, D12S391 and D13S317 exhibited distinct population-specific "most frequent genotypes."

Conclusions: Three different prediction models have been proposed for genotype to geolocation prediction, i.e. (i) use of unique genotypes of a population, (ii) use of the most frequent genotype, and (iii) a combinatorial approach of unique and most frequent genotypes. These models could aid the investigating agencies in cases where no reference sample is available for comparison of the profile.