Transferring a microhaplotype-specific probabilistic genotyping model from targeted amplification to hybridization...

Probabilistic genotyping (PG) has become the standard framework for evaluating forensic DNA mixtures, yet most implementations were developed for STR data and PCR-based enrichment. Microhaplotypes (MHs) provide high allele diversity without stutter and are attractive for mixture deconvolution, particularly when combined with hybridization capture for degraded or limited DNA. Here, we developed a 100-locus hybridization capture MH panel and assessed its performance for mixture interpretation using two continuous PG models: (i) a MH-specific Truncated Gaussian (TG) model previously developed by our group for targeted amplification MH-MPS data, and (ii) the gamma-based model implemented in EuroForMix (EFM). Panel performance was first examined using sensitivity (0.5, 0.125, and 0.0625 ng; triplicates) and repeatability/consistency (10 individuals, 0.5 ng; duplicates) experiments. We then analyzed two- and three-person mixtures with increasing mixture imbalance (2-person: 1:1-1:40; 3-person: 1:1.5:3-1:4:20; triplicates). Across all mixtures, true contributors yielded likelihood ratios (LRs) > 1 under both models, with all tested non-contributors yielding LRs < 1. For two-person mixtures, major-contributor deconvolution accuracy approached 100%, whereas minor-contributor accuracy peaked at 83% (1:5) and declined with increasingly imbalanced mixtures. In three-person mixtures, minor-contributor accuracy depended strongly on relative proportions among contributors and showed non-monotonic trends. Overall, the TG model produced higher LRs and improved deconvolution for minor contributors compared with EFM, supporting transferability of the TG framework from targeted amplification to hybridization capture MH-MPS data. These results provide a preliminary practical foundation for probabilistic interpretation of hybridization capture microhaplotypes in forensic mixtures.