Fecal source identification using random forest.

Area of Science:

• Microbiology
• Environmental Science
• Bioinformatics

Background:

• Gut bacteria like Clostridiales and Bacteroidales show host-specific patterns, making them potential indicators of fecal contamination.
• Accurate identification of fecal pollution sources is vital for managing waterborne diseases and assessing environmental risks.
• Existing computational methods for analyzing these bacterial markers are limited.

Purpose of the Study:

• To develop and evaluate a computational approach using Clostridiales and Bacteroidales for identifying fecal pollution sources.
• To assess the accuracy, consistency, and sensitivity of this method in various sample types.
• To provide a tool for effective risk assessment and management of water contamination.

Main Methods:

• Utilized random forest algorithm with 16S rRNA gene amplicon sequences.
• Focused on bacterial sequences from Clostridiales and Bacteroidales orders.
• Benchmarked classification accuracy using fecal, environmental, and in silico generated samples.

Main Results:

• Classifiers accurately identified human and animal fecal sources (approx. 90% accuracy).
• Random forest predictions demonstrated high reproducibility and consistency.
• The method detected low concentrations of fecal signatures (as low as 0.5% in silico).

Conclusions:

• Random forest classification of 16S rRNA gene amplicons provides a sensitive and accurate method for fecal source tracking.
• This approach effectively identifies microbial signatures of human and animal waste in environmental samples.
• The developed method aids in rapid detection and management of fecal contamination in water resources.