Abstract
BACKGROUND: Identifying key characteristics of unknown suspects, such as age, height, and drug use, is essential for advancing forensic investigations. METHODS: In this study, we employed full-length 16S rRNA gene sequencing to analyze the bacterial communities of nasal skin (NaS), oral epithelial skin (OrE), and palm skin (PaS) in heroin users and healthy controls. RESULTS: Our results revealed a significant reduction in bacterial community diversity among heroin users compared to the control group. Notably, bacterial composition differences were more pronounced in OrE and PaS than in NaS, with genus-level variations being more significant than those at the phylum level. Differential bacterial taxa were identified across all three distinct skin sites, with the most remarkable differences observed in OrE samples. In both NaS and OrE, the differential bacterial genera between the two groups were predominantly aerobic. In addition, the Random Forest model constructed based on the RFE feature selection strategy demonstrated strong potential of OrE bacteria for distinguishing heroin users from non-users, achieving an optimal classification accuracy of over 94% and AUC values exceeding 0.988. CONCLUSION: In conclusion, this study highlights the differences in bacterial diversity between heroin users and healthy controls across different skin sites, supporting the potential use of skin bacterial features as forensic identifiers for heroin use. Future research should explore the mechanisms underlying these microbiome alterations and incorporate larger sample sizes to enhance the robustness of findings.