Microbial Community Dissimilarity for Source Tracking with Application in Forensic Studies

Microbial source-tracking is a useful tool for trace evidence analysis in Forensics. Community-wide massively parallel sequencing profiles can bypass the need for satellite microbes or marker sets which are unreliable when handling unstable samples. We propose a method utilizing Aitchison distance to select important suspects/sources and then propose several methods to estimate the proportions of microbial communities/samples coming from important suspects/sources. A series of comprehensive simulation studies show that these methods are capable of accurate selection and improve the performance of current methods Bayesian SourceTracker and FEAST in the presence of noise microbial sources.