Influence of Library Composition on Source Tracker Predictions for Community-Based Microbial Source Tracking

Community-based microbial source tracking (MST) utilizes high-throughput DNA sequencing to profile and compare the microbial communities in different fecal sources and environmental samples. Source Tracker, a program that compares a library of OTUs from fecal sources (i.e., sources) to those in environmental samples (i.e., sinks) in order to determine sources of fecal contamination, is an emerging tool for community-based MST studies. In this study, we investigated the ability of Source Tracker to determine sources of known fecal contamination in spiked, in situ mesocosms containing different source contributors. We also evaluated how Source Tracker results were impacted by accounting for autochthonous taxa present in the sink environment. While Source Tracker was able to predict most sources present in the in situ mesocosms, fecal source library composition substantially influenced the program's ability to predict source contributions. Moreover, prediction results were most reliable when the library contained only known sources, autochthonous taxa were accounted for and when source profiles had low intragroup variability. Although Source Tracker struggled to differentiate between sources with similar bacterial community structures, it was able to consistently identify abundant and expected sources, suggesting that the Source Tracker program can be a useful tool for community-based MST studies.