Comparing Microbial Source Tracking Methods for Precision and Reliability

Abstract

Microbiome research has experienced remarkable growth in recent years, driven by advances in DNA or RNA sequencing technologies and our deepening understanding of the critical roles that microbiota play in diverse ecosystems, including the human body. The term "microbiome" refers to the entire collection of microorganisms in a given environment, whether that is the gut of a human, the soil in a forest, or the water in an ocean. These microorganisms encompass bacteria, viruses, fungi, and other microbes, all of which collectively influence the health and functioning of their host ecosystem. The complexity and diversity of microbiome data have led to the development of various decomposition methods, each tailored to tackle the specific challenges posed by different environments and research objectives. Numerous tools have been developed to estimate the proportion of different contamination sources within a mixture. In this study, we evaluate the accuracy of various source tracking methods using datasets from microbiome studies. In addition to assessing source tracking methods, we also incorporate two widely used cell type deconvolution methods, namely EPIC and PREDE, which are designed to identify missing cell types in a given dataset. Furthermore, we investigate the effectiveness of combined methods by integrating RAD, a source tracking method aimed at filtering out unimportant sources, with either EPIC or PREDE for enhanced accuracy in both source tracking and cell type deconvolution. This research represents a pioneering effort to examine the application of cell type deconvolution methods in source tracking and vice versa. Particularly noteworthy is our focus on scenarios involving missing sources or cell types in the reference data, shedding light on the intricate interplay between these two analytical domains.