Is he really my relative? Kinship cases in substructured populations: a study on the Lebanese population.

Abstract

The probabilistic analysis performed in DNA kinship cases is affected by factors such as the number of loci analyzed, the STR allelic relative frequency dataset, and the population genetic characteristics of the individuals involved in the analysis. The Lebanese population shows high rates of endogamy and consanguinity, which challenges the power of discrimination of DNA profiles and is a potential source of misleading conclusions. The present study aims to evaluate the probability of false positives in relationship testing, especially siblingship, avuncular, and grandparent-grandchildren relationships, when cases are subject to different constraints, including DNA profile sizes (15 vs. 23 STR loci) and inbreeding coefficients (F _ST). A total of 126 794 kinship comparisons were performed for each type of kinship test using a Lebanese database of 505 unrelated healthy individuals. To do so, DNA analysis results were assessed under a pair of propositions involving the stated relationship link between the two individuals vs. unrelatedness, offering quantification through the likelihood ratio (LR), a widely accepted metric to assess the value of scientific results. Among the simulated cases, false associations for full siblingship relationships were observed in 1 995 (1.6%) instances using a DNA profile size of 15 STRs (LR > 1, or log(LR) > 0). False associations decreased to 340 (0.3%) cases when the DNA profile size was increased to 23 STRs, and to 299 (0.2%) and 91 (0.07%) cases using the co-ancestry coefficient for the Lebanese population (0.002) and that recommended by the National Research Committee (0.02), respectively. Simulations of half siblingships (representing half-siblingship, avuncular, and grandparent-grandchildren relationships) showed similar results. Conversely, 41 full biological siblingship cases, 24 true full-siblingship laboratory cases, and 124 true half-siblingship laboratory cases were used as controls to assess the effects of implementing the conditions applied in the simulations (profile size and co-ancestry coefficient (F _ST) values) to lower false associations. Results for the positive controls showed that the implementation of the National Research Committee-recommended F _ST value of 0.02 decreased the LR by an average of 10² for 15-STR DNA profiles, and by an average of 10⁴ for 23-STR DNA profiles, which did not lead to misleading results. However, the use of a DNA profile size of 23 STRs with the Lebanese co-ancestry coefficient showed the most meaningful results. This study could help to set national recommendations for the interpretation of kinship testing in the Lebanese population.