DNA shotgun sequencing evidence: sample-specific and unknown genotyping error probabilities

Abstract: DNA shotgun sequencing evidence is starting to gain a lot of attraction in forensic genetics. Methods to correctly interpret such evidence, including properly accounting for sequencing errors, are needed. This paper extends the wgsLR model by Andersen et. al. (2025) from only allowing for the same, known genotyping error probability for the two samples (trace sample from unknown donor and reference sample from person of interest), to allowing for different genotyping error probabilities (e.g., from trace hair sample and buccal swab reference sample). The model was also extended to be able to integrate out unknown genotyping error probabilities if only a prior probability is known. The sensitivity of the model against overdispersion was also investigated and it was found that it is very robust against overdispersion in estimating the genotyping error probability. It was also found that integrating out unknown genotyping error probability of the trace sample gave concordant weight of evidence under both the hypotheses (first hypothesis that the same individual was the donor of both trace and reference sample as well as the second hypothesis that two different individuals were the donors for the trace and reference sample). It was found that it is more consevative to use prior distributions with a too small mean rather than a too high mean. The extension of the model is implemented in the R package wgsLR.