The interpretation of DNA profiles derived from mixtures and low-template or degraded samples is challenging due to difficulties in estimating mixture proportions and genotype combinations. Probabilistic genotyping software based on fully continuous models is useful for interpreting such profiles. In this study, we developed a new probabilistic genotyping software, KongohPlus, based on maximum likelihood estimation (MLE). In our previous work, KongohPlus employed a Markov chain Monte Carlo (MCMC) approach. In the present study, the estimation framework was revised to MLE to improve computational stability and reproducibility.

KongohPlus is an R Shiny–based package that provides an interactive graphical user interface. MLE is performed using the Covariance Matrix Adaptation Evolution Strategy (CMA-ES), implemented via the R package cmaes. The case-specific parameters include template amounts and degradation parameters for each contributor, as well as amplification efficiencies for each PCR replicate. The laboratory-specific parameters include locus-specific amplification efficiencies, allele-specific stutter ratios (back, forward, double-back, and minus 2-nt stutters), and the scale parameters of the gamma distribution used to model peak heights for alleles and stutters. These laboratory-specific parameters are pre-determined based on user-generated experimental data. The MLE procedure is repeated up to 10 times and is terminated when the same maximum likelihood is obtained in three runs.

We validated KongohPlus using 2–4 person mixtures and degraded DNA profiles from the PROVEDIt dataset (GlobalFiler, 29 cycles, 15 s). The likelihood ratios (LRs) for true contributors and non-contributors were well separated under higher information conditions (i.e., larger peak heights and fewer contributors), whereas they approached neutrality as the information decreased. The LR values were also highly reproducible, and the accuracy of the calculation was also confirmed. These results demonstrate that KongohPlus is useful for interpreting DNA profiles derived from mixtures and low-template or degraded samples.

Authors

• Sho Manabe (Department of Legal Medicine, Kansai Medical University, Japan)
• Masaki Hashiyada (Department of Legal Medicine, Kansai Medical University, Japan)
• Atsushi Akane (Department of Legal Medicine, Kansai Medical University, Japan)