# SourceApp: A Novel Metagenomic Source Tracking Tool that can Distinguish between Fecal Microbiomes Using Genome-To-Source Associations Benchmarked Against Mixed Input Spike-In Mesocosms

## TL;DR

SourceApp is a new tool that uses metagenomics to track and identify sources of fecal pollution in water samples.

## Contribution

SourceApp introduces a novel method for fecal source tracking using genome-to-source associations and is benchmarked using controlled mesocosm experiments.

## Key findings

- SourceApp achieved sensitivity and specificity near 0.90 overall after parameter tuning.
- Cows were underestimated due to limited sampling of cow fecal microbiomes.
- The tool outperformed existing alternatives in source attribution and apportionment.

## Abstract

Methodologies utilizing metagenomics are attractive to
fecal source
tracking (FST) aims for assessing the presence and proportions of
various fecal inputs simultaneously. Yet, compared to established
culture- or PCR-based techniques, metagenomic approaches for these
purposes are rarely benchmarked or contextualized for practice. We
performed shotgun sequencing experiments (n = 35)
of mesocosms constructed from the water of a well-studied recreational
and drinking water reservoir spiked with various fecal (n = 6 animal
sources, 3 wastewater sources, and 1 septage source) and synthetic
microbiome spike-ins (n = 1) introduced at predetermined
cell concentrations to simulate fecal pollution events of known composition.
We built source-associated genome databases using publicly available
reference genomes and metagenome assembled genomes (MAGs) recovered
from short- and long-read sequencing of the fecal spike-ins, and then
created an associated bioinformatic tool, called SourceApp, for inferring
source attribution and apportionment by mapping the metagenomic data
to these genome databases. SourceApp’s performance varied substantially
by source, with cows being underestimated due to under sampling of
cow fecal microbiomes. Parameter tuning revealed sensitivity and specificity
near 0.90 overall, which exceeded all alternative tools. SourceApp
can assist researchers with analyzing and interpreting shotgun sequencing
data and developing standard operating procedures on the frontiers
of metagenomic FST.