The human microbiome presents a promising new method for forensic body fluid inference, complementing other molecular approaches such as the use of mRNA and methylation markers. Its utility for the investigation of sexual assault cases has been highlighted in several recent studies demonstrating that sexually shared bacteria can be identified in vaginal and penile skin swabs collected shortly after sexual activity. These bacteria have the potential to assist in reconstructing criminal activities and to support donor matching, particularly when using long-read sequencing data.  While these findings are promising, further studies are needed to evaluate the persistence and robustness of microbial signatures of sexual contact over time. In addition, machine learning approaches provide a valuable framework for evaluating microbiome-based forensic methods. In this study we explored genital microbiome composition before and after sexual intercourse in 13 heterosexual couples, examining persistence by focusing on four post-coital time points (immediately, 24, 48, and 72 hours after intercourse). In addition to vaginal and penile skin swabs, we analysed saliva, semen, and fingernail samples. Female participants measured vaginal pH, and male participants provided underwear samples. Detailed metadata on health and behaviour were collected through questionnaires. Full-length 16S rRNA gene sequencing was performed using the PacBio Kinnex library preparation approach and sequencing on the Revio platform. Read processing and statistical analyses were conducted using QIIME2 and suitable packages in R, to explore within- and between-sample diversity, species-level taxonomic composition, and microbial sharing patterns between partners. A machine learning classifier was applied to evaluate body fluid classification performance for both pre- and post-coital samples. Our results demonstrate pronounced and relatively persistent shifts in both the vaginal microbiome and penile microbiome following sexual activity, even when barrier contraception was used. Vaginal microbiota changes were reflected in CST shifts and were generally accompanied by increases in vaginal pH. We also observed an increased number of shared taxa between partners immediately after intercourse, with a decline over time. In addition, we identified taxa uniquely shared between partners, including Gardnerella species, underscoring their potential utility for donor matching. Several of these taxa (e.g. Gardnerella) are associated with bacterial vaginosis, highlighting the clinical relevance of partner-mediated microbial transfer. Machine learning classification showed generally high accuracy for post-coital samples, although some limitations were observed. These findings advance our understanding of microbial dynamics associated with sexual activity, and highlight both the potential and current limitations of microbiome-based forensic inferences.

Authors

• Natasha Arora (Forensic Genetics Department, Zurich University of Forensic Medicine, University of Zurich, Switzerland)
• Jasmin Baumgartner (Forensic Genetics Department, Zurich University of Forensic Medicine, University of Zurich, Switzerland)
• Meghna Swayambhu (Forensic Genetics Department, Zurich University of Forensic Medicine, University of Zurich, Switzerland)
• Peter Resutik (Forensic Genetics Department, Zurich University of Forensic Medicine, University of Zurich, Switzerland)
• Monica Ticlla (Department of Research, Swiss Tropical and Public Health Institute, University of Basel, Switzerland)
• Carla Hart (Forensic Genetics Department, Zurich University of Forensic Medicine, University of Zurich, Switzerland)
• Simon Grüter (Functional Genomics Centre Zurich, ETHZ/University of Zurich, Switzerland)
• Anna Bratus-Neuenschwander (Functional Genomics Centre Zurich, ETHZ/University of Zurich, Switzerland)