Evidentiary evaluation of complex low-template DNA mixtures using high-efficiency microhaplotype panels

Low-template DNA mixtures pose substantial analytical challenges owing to limited genetic information, allelic dropout, and increasing complexity arising from multiple contributors and imbalanced mixture proportions. Using highly informative genetic markers in combination with fully continuous probabilistic genotyping (PG) models has been widely recognized as a promising strategy for improving the interpretation of complex DNA mixtures. Accordingly, in this study, we evaluated the evidential performance of low-template DNA mixtures using high-efficiency next-generation sequencing-based microhaplotype (MH) marker systems in combination with PG. Three MH panels (55-, 67-, and 87-plex) were examined under forensically relevant conditions, including low DNA input (down to 0.05 ng), increasing numbers of contributors (up to four), and extreme mixture ratios (up to 1:40). Likelihood ratio (LR) distributions were generated using the EuroForMix software by designating either a true minor contributor or a non-contributor as the person of interest. In two-person balanced mixtures, all panels produced reliable results even at the lowest DNA input, consistently distinguishing true contributors from non-contributors, including close relatives. In complex multi-person mixtures, panels with relatively higher polymorphism improved genotype resolution, reducing spurious LR inflation for relatives and enhancing contributor/non-contributor separation. Under highly imbalanced conditions, panels with higher locus counts partially compensated for allelic information losses and retained limited but informative discriminatory power. These findings indicate that MH panel performance is strongly context-dependent, impacted by mixture composition, locus number, and marker polymorphism. As supplementary analyses, mixture deconvolution accuracy was assessed using a mixture proportion deviation metric (D_ratio), and kinship inference was explored in the absence of direct reference profiles by comparing LR support for close relatives versus unrelated individuals. Overall, these results provide valuable insights for guiding panel design, analytical strategies, and interpretation frameworks for increasingly challenging forensic DNA mixture analysis.