A continuous model for interpreting microhaplotype profiles of forensic DNA mixtures

Microhaplotypes (MHs) have great potential in forensic DNA analysis, with applications in individual identification, kinship analysis and ancestry inference. No matter the forensic application, the analysis of DNA mixtures may be encountered. This study aims to develop and evaluate a continuous model for interpreting mixed genotype data from MH markers. We characterized MH profile features and modeled allele read counts using a truncated Gaussian distribution, accounting for allele dropout, noise, and locus-specific detection efficiency. The model was tested on 90 DNA mixtures generated from nine unrelated individuals across various mixture proportions. Likelihood ratio (LR) values were computed for both true contributors and non-contributors, and mixture deconvolution was performed. Results demonstrated high accuracy and specificity in interpreting MH profiles for 2- to 3-person DNA mixtures: true contributors obtained LR values greater than 1 in 190 out of 200 LR calculations. In 26,700 simulated non-contributor tests, for 2-person mixtures, the proportion of non-contributors with an LR greater than 1 was 0.0051 %; for 3-person mixtures, this proportion was 4.68 %. Excluding balanced individuals in mixtures, the average deconvolution accuracy rate for major contributors was 0.9145, with 60.98 % (100/164) achieving an accuracy rate of 1. Additionally, we observed that distinguishing alleles from non-alleles became increasingly challenging with higher mixture proportions or additional contributors, with noise identified as a critical factor affecting genotyping accuracy.