Comparison of key diagnostics for probabilistic interpretation of STR mixture data generated with length-based and...

This study directly compared the mixture deconvolution performance of two probabilistic genotyping software applications: STRmix™ v2.8, which accepts length-based STR input data, and STRmix™ NGS v1.1.0, which accepts sequence-based STR input data. The same set of two-, three- and four-person DNA mixtures was used to generate data with the length-based GlobalFiler™ kit as well as the massively parallel sequencing (MPS) ForenSeq® Signature Prep and MainstAY kits, and the three resulting data sets were subsequently interpreted with the appropriate probabilistic genotyping software (PGS). The comparison of the software applications focused on three central interpretation diagnostics: likelihood ratio, mixture proportion, and genotype weight. Sequence-based likelihood ratios were on average 5 orders of magnitude higher than their length-based counterparts. Three possible factors that could account for this increase- assay sensitivity, isoallele detection, and isoallele population frequencies- were explored. To examine mixture proportions, template levels were considered, including bead normalization with MPS library preparation methods. The results of the comparisons of template levels, mixture components, and genotype weights indicate that STRmix™ v2.8 and STRmix™ NGS v1.1.0 have similar levels of deconvolution efficacy, meaning that the increased power of discrimination inherent in MPS data can be fully realized in a software-based approach to mixture interpretation without loss of genotyping accuracy. In addition, sequence information can provide invaluable assistance to experts in establishing the foundational user-defined assumptions under which these PGS systems operate. Overall, the study demonstrates that integrating MPS technologies with probabilistic genotyping offers a robust approach for enhancing STR mixture interpretation.