SMART-MHmix: A probabilistic model for microhaplotype-based forensic DNA mixture analysis.

Area of Science:

• Forensic Genetics
• Molecular Biology
• Biostatistics

Background:

• Interpreting complex DNA mixtures is a significant challenge in forensic genetics.
• Capillary electrophoresis (CE)-based short tandem repeat (STR) analysis has limitations with highly complex mixtures.
• Microhaplotypes (MHs) analyzed via next-generation sequencing (NGS) offer a promising alternative but lack quantitative interpretation methods.

Purpose of the Study:

• To introduce SMART-MHmix, a probabilistic genotyping framework for quantitative interpretation of NGS-based MH data.
• To enable statistical evidentiary assessment for complex forensic DNA mixtures.
• To address the gap in analyzing complex MH mixtures using NGS technology.

Main Methods:

• Developed SMART-MHmix, a probabilistic framework modeling mixture profiles from NGS-sequenced MH loci.
• Calculated the probability of observed allele read counts using a model incorporating contributor DNA amounts, amplification efficiencies, and replicate effects.
• Integrated a log-normal distribution for allele signals, models for drop-in/drop-out events, and Markov Chain Monte Carlo (MCMC) sampling.

Main Results:

• SMART-MHmix demonstrated robust performance across 104 synthetic DNA mixtures (2-5 contributors).
• Likelihood ratio (LR) analysis showed strong discrimination, with LRs for true contributors up to 10^47 and >86% exceeding 10^5.
• Sensitivity and specificity exceeded 90% with a threshold of θ=1; accurate genotype inference was achieved for major contributors.

Conclusions:

• SMART-MHmix provides a dedicated probabilistic model for MH-NGS data, enabling reliable analysis of complex forensic DNA mixtures.
• The framework shows potential to outperform conventional CE-STR workflows in forensic applications.
• Accurate inference of major and second contributors was demonstrated, with performance declining for minor contributors in complex mixtures.