Submerged environments, characterized by complex physicochemical and biological variables, frequently facilitate the concealment of criminal evidence. Consequently, research focusing on the estimation of the post-submersion interval (PSI) for submerge...
Submerged environments, characterized by complex physicochemical and biological variables, frequently facilitate the concealment of criminal evidence. Consequently, research focusing on the estimation of the post-submersion interval (PSI) for submerged exhibits is critically warranted. This research assessed the feasibility of PSI estimation for bloodstain evidence exposed to diverse aquatic conditions by integrating water quality parameters, DNA quantification values, STR profiles, and microbiome data using machine learning (ML) methodologies. The methodology involved measuring water quality concurrently with sample collection to monitor temporal chemical change. DNA extraction utilized the QIAamp DNA Mini Kit (QIAGEN, USA) for the substrate and the DNeasy PowerWater Kit (Qiagen, USA) for filter paper. For short-term submersion, STR profiles were obtained from human DNA to construct a PSI algorithm. For long-term analysis, next-generation sequencing (NGS) was performed on 11 filter paper samples with MiSeq, with bioinformatics analysis conducted using QIIME2 and MicrobiomeAnalyst platforms.
The mixed data driven PSI estimation yielded notable predictive performance. Random forest (RF) multi-class classification model achieved a maximum classification accuracy of 77.78%, and the XGBoost regression model recorded a mean absolute error (MAE) of 0.73 hours. Furthermore, prediction accuracy was significantly augmented when physicochemical independent variables such as pH and oxidation–reduction potential (ORP) were incorporated alongside STR data. NGS analysis identified the Proteobacteria phylum as the predominant taxon, exceeding 99% in distilled water (D.W.). While its relative abundance declined in reservoir and seawater over prolonged submersion, it maintained high dominance relative to controls. This research establishes a robust foundation for PSI estimation by STR profiles and water quality variables up to 7 days. Furthermore, the PSI for the late-submersion phase is expected to be estimated using microbial biomarkers. The findings hold substantial potential to evolve into a critical forensic tool, offering significant investigative leads in submerged crime scenarios, contingent upon further algorithmic refinement and expanded analysis to enhance accuracy and generalizability.