Body fluid identification (BFID) and estimation of time since deposition (TsD) are valuable yet challenging in forensic practice. Previous studies have demonstrated that integrating microbial and metabolomic profiles provides complementary biological insights. Therefore, this study performed untargeted metabolomic profiling and full-length 16S rRNA sequencing on fresh saliva (SA), semen (SE), vaginal secretions (VF), and their mixtures (SA-VF and SE-VF), with additional microbial analysis after 15 and 30 days of indoor exposure. Results showed the single-fluid samples exhibited specific dominant bacterial taxa, whereas the two mixture samples contained detectable bacterial signatures from both constituent fluids. Untargeted UHPLC-QTOF/MS analysis revealed unique metabolic signatures for each body fluid, enriched in biologically relevant pathways like steroid and bile acid metabolism. Moreover, we putatively identified characteristic metabolites, including α-solanine, candicidin, and megalomicin C1, some of which are rare microbial antibiotics. Owing to the exploratory nature and associated constraints of nontargeted approaches, these results serve as a provisional reference for identifying potential candidates. Integration of metabolomic and microbiome data uncovered strong metabolite-microbe correlations, highlighting microbially influenced metabolic networks unique to each body fluid type. Using differential microbes and metabolites individually as input features, the random forest model achieved BFID accuracies of 80 and 83.1%, respectively; however, integrating both sets of features increased accuracy to 100%. In contrast, microbial-based TsD prediction performed well for single-fluid samples but showed reduced effectiveness for mixed samples. Overall, our research highlights the powerful predictive potential and improved predictive accuracy of the integration of microbiome and metabolome data in BFID.
Deng et al. (Mon,) studied this question.