Age prediction using DNA methylation has become an important tool in forensic genetics, especially for semen samples in sexual assault cases. Our previous research has identified age-associated CpG markers on the Y-chromosome and developed a multiplex assay along with regression models for predicting the age of semen samples. However, the performance of these methods under biologically altered conditions, such as vasectomized semen and mixed body fluid samples, has not been thoroughly investigated. In this study, we analyzed semen samples from vasectomized individuals using a previously developed 10-amplicon Y-CpG assay to examine methylation patterns and age prediction capabilities. The vasectomized samples were categorized based on body fluid identification results, revealing distinct methylation profiles. We observed markedly reduced age prediction accuracy in samples that lacked semen-specific profiles, while those with semen-specific profiles exhibited patterns similar to normal semen. Building on feature selection outcomes from our earlier study, we further developed an optimized five-amplicon multiplex assay and assessed its performance with independent semen samples and semen-vaginal fluid mixtures. In evaluations involving 11 independent semen samples, the stepwise and lasso models produced mean absolute errors (MAEs) of 6.74 and 6.55 years, respectively. The assay also showed consistent performance with semen-vaginal fluid mixtures, with methylation differences below 2% and average age prediction errors under 6 years. Overall, these findings indicate that the five-amplicon assay offers a streamlined and practical approach for predicting semen age in forensic contexts. Additionally, incorporating body fluid identification may enhance the interpretation of age predictions in challenging samples.
Lee et al. (Tue,) studied this question.