Objectives

Accurate estimation of the post-mortem interval (PMI) during summer is challenging due to accelerated decomposition and intense entomological activity, which often compromise traditional indicators. While volatile organic compounds (VOCs) act as early, readily detectable chemical signals driven by microbial metabolism, their integrated value for PMI estimation remains underexploited. This study aims to develop a robust and interpretable PMI estimation framework by integrating odor-active VOCs and key bacterial taxa. By leveraging machine learning, we seek to identify optimal, mechanistically validated forensic biomarkers for precise early PMI estimation under summer conditions.

Material and methods

A decomposition model comprising 40 male Sprague-Dawley rats was deployed across three distinct summer environments: indoor (insect-excluded), outdoor, and buried. Oral microbiome swabs and VOC twisters were sampled at 0 h, 12 h, 24 h, and 72 h post-mortem. VOC profiles were characterized utilizing thermal desorption unit-gas chromatography-mass spectrometry (TDU-GC-MS) and gas chromatography-olfactometry (GC-O), while microbial succession was analyzed via 5R 16S rRNA gene sequencing. Spearman correlation analysis, random forest (RF) regression, and orthogonal partial least squares discriminant analysis (OPLS-DA) were applied to identify VOCs significantly associated with PMI. Crucially, key bacterial candidates strongly correlated with specific VOCs (r > 0.6, p-adjust < 0.05) were cultured in vitro to biologically validate their VOC production capabilities.

Results and conclusions

Principal Coordinate Analysis (PCoA) revealed that VOC profiles exhibited more distinct temporal clustering than microbial communities during decomposition. RF regression models based solely on VOCs outperformed microbiome-only models (R²=0.94, MAE=5.76 hours vs. R²=0.85, MAE=9.44 hours). Notably, integrating both multi-omics signatures achieved comparable high accuracy (R²=0.96, MAE=3.32 hours, RMSE=5.11 hours). Feature selection identified Dimethyl disulfide (DMDS) and Dimethyl trisulfide (DMTS) as highly significant PMI predictors, which strongly correlated with Proteus mirabilis, Enterococcus faecalis, and Myroides odoratimimus. Subsequent in vitro assays successfully confirmed these targeted bacteria as the biological source of the VOCs. Ultimately, a streamlined, field-applicable model utilizing only these three bacterial species and two VOCs maximized predictive accuracy (R²=0.98, MAE=2.61 hours, RMSE =3.24 hours). These findings demonstrate that integrating microbiome and VOC data provides robust, complementary mechanistic insights. The proposed framework not only offers high predictive performance across diverse environments but also establishes a biologically validated, interpretable tool with significant potential for early PMI estimation in challenging summer scenes.

Authors

• Ji Chen (Institute of Forensic Science, Fudan University; School of Forensic Medicine and Science, Fudan University, China)
• Xiang Wang (Institute of Forensic Science, Fudan University; School of Forensic Medicine and Science, Fudan University, China)
• Chengtao Li (Institute of Forensic Science, Fudan University; School of Forensic Medicine and Science, Fudan University, China)
• Suhua Zhang (Institute of Forensic Science, Fudan University; School of Forensic Medicine and Science, Fudan University, China)