U.S. flag

An official website of the United States government, Department of Justice.

NCJRS Virtual Library

The Virtual Library houses over 235,000 criminal justice resources, including all known OJP works.
Click here to search the NCJRS Virtual Library

Microbiome Data Accurately Predicts the Postmortem Interval Using Machine Learning Regression Models

NCJ Number
255741
Journal
Genes Volume: 9 Issue: 2 Dated: 2018
Author(s)
Aeriel Belk; Zhenjiang Zech; Xu Zhenjiang; Zech Xu; David O. Carter; Aaron Lynne; Sibyl Bucheli; Rob Knight; Jessica L. Metcalf
Date Published
February 2018
Annotation

In this publication, researchers demonstrate that microbiome data can accurately predict the postmortem interval (PMI) using machine learning models.

Abstract

In this study, the authors explore how to build the most robust Random Forest regression models for prediction of postmortem interval (PMI) by testing models built on different sample types (gravesoil, skin of the torso, skin of the head), gene markers (16S ribosomal RNA (rRNA), 18S rRNA, internal transcribed spacer regions (ITS)), and taxonomic levels (sequence variants, species, genus, etc.). The authors also tested whether particular suites of indicator microbes were informative across different datasets. Generally, results indicate that the most accurate models for predicting PMI were built using gravesoil and skin data using the 16S rRNA genetic marker at the taxonomic level of phyla. Additionally, several phyla consistently contributed highly to model accuracy and may be candidate indicators of PMI. Death investigations often include an effort to establish the postmortem interval (PMI) in cases in which the time of death is uncertain. The PMI can lead to the identification of the deceased and the validation of witness statements and suspect alibis. Recent research has demonstrated that microbes provide an accurate clock that starts at death and relies on ecological change in the microbial communities that normally inhabit a body and its surrounding environment. (Published Abstract Provided)