Integrating microbial profiling and machine learning for inference of drowning sites: a forensic investigation in the Northwest River.
Journal:
Microbiology spectrum
PMID:
39651862
Abstract
Drowning incidents present significant challenges for forensic investigators in determining the exact site of occurrence. Traditional forensic methods often rely on physical evidence and circumstantial clues, but the emerging field of forensic microbiology offers a promising avenue for enhancing precision and reliability in site inference. Our study investigates the application of microbiome analysis in inferring drowning sites, focusing on microbial diversity in water samples and lung tissues of drowned animals from different sites in the Northwest River. We utilized 16S rDNA sequencing to analyze microbial diversity in water samples and lung tissues, revealing distinct microbial signatures associated with drowning sites. Our findings highlight variations in species richness and diversity across different sampling points, indicating the influence of environmental factors on microbial community structure. Machine learning models trained on microbial data from lung tissues demonstrated high accuracy in predicting drowning sites, with cross-validation accuracy ranging from 83.53% ± 3.99% to 95.07% ± 3.17%. Notably, the Gradient Boosting Machine (GBM) method achieved a classification accuracy of 95.07% ± 3.17% for different sampling points at a submersion time of 1 day. Moreover, our cross-species site inference results revealed that utilizing data from drowned mice to predict the drowning sites of rabbits in location W5 achieved an accuracy of 72.22%. In conclusion, our study underscores the potential of microbiome analysis in forensic investigations of drowning incidents. By integrating microbial data with traditional forensic techniques, there is significant potential to enhance the reliability of scene inferences, thereby making substantial contributions to case investigations and judicial trials.IMPORTANCEBy employing advanced techniques like microbial profiling and machine learning, the study aims to enhance the accuracy of determining drowning sites, which is crucial for both legal proceedings. By analyzing microbial diversity in water samples and drowned animal lung tissues, the study sheds light on how environmental factors and victim-related variables influence microbial communities. The findings not only advance our understanding of forensic microbiology but also offer practical implications for improving investigative techniques in cases of drowning.