Y-chromosomal kinship estimation for forensic familial searching: YMrCA to the rescue

Investigative Genetic Genealogy (IGG) is an identification method in forensic case work where instead of finding the donor of a trace, a relative of the donor is used to unravel the identity of the unknown perpetrator. As autosomal DNA kinships fade away over generations due to recombination events, Y-chromosome analysis provides the opportunity to identify distant kinships as 95% of the chromosome lacks recombination. A close or distant biological relative can be verified using (rapidly mutating) short tandem repeats or (RM) Y-STRs, which identifies familial Y-haplotypes []. Through Y-haplotype comparison, the time to their most recent common ancestor (tMRCA) needs to be calculated. The more accurate the number of generations is estimated, the easier it becomes to reconstruct the genealogy. In a previous study, we investigated the two state-of-the-art mutation models (SMM: stepwise mutation model; IAM: infinite alleles model) and three online tMRCA calculators (McDonald, Walker and McGee) []. After running our 1120 biologically related genealogical pairs, however, consistent under- and overestimation and broad confidence intervals were observed, leading to dubious tMRCA estimates. This mainly because they do not include individual mutation rates with multi-step changes, while ignoring hidden multiple, back or parallel modifications. To improve tMRCA estimation, we developed a user-friendly calculator, the ‘YMrCA’, including all previously mentioned mutation characteristics []. In this study, a specific report case using genealogical pairs with confirmed biological kinships are presented to visualize the performance of the YMrCA compared to the state-of-the-art.

Two unique genealogical pairs with both four Y-STR changes on 46 Y-STR haplotype, but separated by different number of generations, were used to estimate their kinship. Pair A is separated by 10 generations and has three one-step changes (DYS499, DYS481 and DYS627) and one two-step change (DYS459). Pair B is separated by 20 generations having four 1-step differences (DYS385, DYS456, DYS518 and DYS627).

Fig. 1 provides an overview of the results below. The SMM estimated for couple A 16 generations and for B 13 generations. This difference is because the model counted five mutation events for couple A due to its two-step mutation event. The IAM estimated for both couples 13 generations as they have equal number of Y-STR differences. This revealed a generation deviation for couple A of 6 generations (SMM) and 3 generations (IAM), while couple B resulted for both models in a generation deviation error of 7 generations. For the online tMRCA calculators, both the McDonald as Walker calculator estimated 16 generations for couple A and B with a 95% confidence interval of 3–40 generations. This resulted in a generation error of 6 generations for couple A and 4 generations for couple B. Without the inclusion of exact Y-STR allele calls or individual mutation rates, these calculators estimate equal probability curves for genealogical pairs encountering equal number of Y-STR differences. The McGee calculator was, despite the more advanced input of allele calls and their use of individual mutation rates, very discordant with reality. McGee estimated for couple A 36 generation (error +26) and for couple B 29 generations (error +9). This calculator did not provide probability curves but stated that the probability is 95% that the tMRCA is no longer than the indicated generation output. So in general, the state-of-the-art calculators estimated broad confidence intervals and high generation deviations, decreasing the tMRCA estimation accuracy which is disadvantageous for pedigree reconstruction and kinship analysis.

The YMrCA calculator, on the other hand, resulted in exact the same number of generations for the two couples with a significantly smaller 95% confidence interval for couple A (Fig. 1, right). The adaptation of the input Y-STR matrices for the YMrCA calculator with these haplogroup-specific mutation rates increased estimation probabilities for the expected tMRCA and decreased both generation estimation errors towards a more truthful tMRCA prediction. To conclude, the overall high success rate of the YMrCA and its low deviation error constitute promising results compared to both the basic SMM and IAM, as to the already existing online tMRCA calculators. Additionally, the incorporation of individual Y-STR mutation rates and even haplogroup-specific ones into the YMrCA provides improved tMRCA output information for genealogical pairs having equal number of Y-STR differences.

Here we demonstrate that the mutation models and online calculators are not appropriate for use in forensic familial searching due to their large 95% CI, making predictions too difficult for pedigree reconstructions, and the consistent tMRCA under- and overestimations, leading to implausible and dubious probability scores. Our newly developed YMrCA demonstrated promising performance, with a significantly higher tMRCA estimation success rate following reality due to the inclusion of individual Y-STR mutation rates and characteristics.

We thank all DNA donors for their participation in our genetic-genealogy project. Funding was provided by FWO (SC: 1265622N).