Probabilistic genotyping of non-human DNA mixtures: using MaSTR™ to analyze mixed canine STRs genotyped with DogFiler

Individualization in forensic science is commonly achieved by typing short tandem repeats (STRs). Interpretation of resulting electropherograms can be challenging when the input DNA was low quantity, low quality or a mixture of two or more individuals. Probabilistic genotyping (PG) – application of mathematical models and algorithms to analyze mixed STR profiles – tools have been developed to compute a likelihood ratio that an individual is a contributor, and/or deconvolute the mixed profile into single source genotypes. While PG software has been widely implemented in human forensic laboratories, wildlife forensic laboratories have been slow to adopt given a) separate validations would be needed for each STR panel in use, and b) access to and selection of relevant population data needed for analysis could be challenging. This study looked to complete the first evaluation of PG software for the analysis of non-human STR mixtures. We utilized a continuous PG software, MaSTR™ (SoftGenetics), that permits the creation of a model for a non-human STR panel. Artificial mixtures of domestic dog were prepared at varied ratios and genotyped with the published DogFiler STR panel: 1) unrelated two individual (n = 12) and three individual (n = 3) mixtures, with varying degrees of allele sharing, and 2) related individuals (mother and offspring; n = 2) with high allele sharing. For each mixture, true contributor single source profiles were provided to MaSTR™ under varying propositions to assess the impact on the computed likelihood ratio (LR). Additionally for a subset of unrelated two-individual mixtures (n = 6), analyses with true non-contributors were completed to assess false inclusion rates. For replicate analyses (n = 5) on the same mixed profile, the coefficient of variation in the LogLR was 1.83 % and 2.42 %, for two-and three-individual mixtures, respectively. The overall false exclusion rate of true contributors was 0.43 % (LogLR < 6), and the false inclusion rate with true non-contributors was 0 % (LogLR > 6). Despite nearly half of the mixtures tested consisting of individuals with high levels of allele sharing, MaSTR™ was able to complete the analysis and give the expected result (inclusion or exclusion) regardless of the mixture ratio. These results provide promise for use of MaSTR™ (and PG more generally) in wildlife forensic casework. Laboratories interested in implementing PG will need to complete necessary validations and ideally include in testing STR profiles generated from adjudicated case samples, which was outside the scope of this study.