Cell-type-dependent expression of a somatic tri-allelic pattern at D18S51 and Penta D: The statistical interpretation of forensic evidence in sexual assault investigations

Abstract

Tri-allelic patterns, though rare, present a significant interpretive challenge in forensic genetics. Their misinterpretation as a DNA mixture can lead to flawed statistical assessment, particularly for likelihood ratio (LR) calculations. However, when correctly identified, a tri-allelic pattern can serve as a powerful identifying marker, increasing the evidential value of a match between a POI (person of interest) and forensic evidence. This case report analyses a sexual assault investigation involving the analysis of cloth fragments, subungual, anal, vulvar, and vaginal swabs, alongside reference samples from the victim and suspect. The POI’s buccal reference sample exhibited a tri-allelic pattern at the D18S51 locus (alleles: 16, 17, 18) and a substantial peak height imbalance (PHR = 0.16) at Penta D. We found that this tri-allelic pattern corresponds to a Type 1, an early embryonic somatic mutation. While some evidence items yielded mixed profiles of the victim and suspect, others including a cloth fragment, two subungual swabs, and an anal swab revealed a single-source male profile matching the POI. Critically, the tri-allelic pattern was observed across multiple evidence types, but the relative proportion of the three alleles (the intra-locus ratio) varied systematically depending on the tissue origin of the sample. This finding suggests cell-type-dependent genetic makeup of the somatic mutation, a phenomenon known as mosaicism. We discuss the characteristics of this pattern, postulate its embryonic origin, and propose a framework for its statistical incorporation into the LR. This case underscores that recognition of the biological mechanism is essential to avoid misinterpretation and can provide stronger evidence of genetic identification in forensic casework. The freely available app LRtri, https://francomarsico.shinyapps.io/LRtri/, can be used to explore the uncertainty in the LR.