The Microbiome Forensics Database UZH. (December 2022)

• Record Type: Journal Article
• Authors: Arora, Natasha; Matias Rodrigues, João F.; Swayambhu, Meghna; Witlox, Pim
• Is Part Of: Forensic science international. Volume 8(2022)
• Page Start: 181
• Page End: 184
• Publication Date: 2022-12
• Subjects: Microbiome forensics -- Body fluid/tissue identification -- Training dataset -- Machine learning -- Database; Forensic genetics -- Periodicals; Forensic Genetics -- Periodicals; Electronic journals; 614.1
• DOI: 10.1016/j.fsigss.2022.10.028
• Languages: English
• ISSNs: 1875-1768

Abstract

Abstract: Microbial communities in biological stains can provide valuable information to assist forensic scientists identify the body fluid/tissue present in these. As these microbial communities are characteristic of body habitats, DNA sequencing of microbes can be used to predict bodily origin. Promising predictive results have been obtained with supervised machine learning algorithms trained on bacterial abundance data from human body sites. Importantly, prediction accuracy is dependent on the training dataset, yet compiling a large and comprehensive training reference is a non-trivial issue requiring substantial efforts. Here we present a new online database and associated data-mining platform which is, to our knowledge, the first one customised for forensic scientists investigating body fluids/tissues. Our database features samples originating from ten human body sites, with selection options through an online platform. Users can download bacterial abundance as well as taxonomic data, which can then be used to train predictive models and test their accuracy. Future stages of the development of the platform will include curation of the samples to decrease potential errors in sample labelling, as well as access to an online tool to conduct exploratory analyses.