SourceFinder: a Machine-Learning-Based Tool for Identification of Chromosomal, Plasmid, and Bacteriophage Sequences...

Area of Science:

• Genomics
• Bioinformatics
• Microbial Ecology

Background:

• High-throughput sequencing facilitates studying genetic interactions like horizontal gene transfer.
• Identifying plasmids and bacteriophages from assembled reads is difficult due to genome plasticity and assembly challenges.

Purpose of the Study:

• To develop a robust classifier for identifying the origin of bacterial DNA sequences (chromosome, plasmid, or bacteriophage).
• To provide a tool that assists researchers in analyzing bacterial genome data and understanding gene dissemination.

Main Methods:

• A random forest classifier was trained on 23,211 complete bacterial, plasmid, and bacteriophage sequences.
• Sequences were subsampled into fragments and subdivided into k-mers to adapt the classifier for incomplete data.
• The classifier's performance was evaluated using k-mer distributions on complete and partial genome sequences.

Main Results:

• The SourceFinder classifier accurately identified chromosomal, plasmid, and bacteriophage sequences.
• Performance metrics included a minimum area under the receiver operating characteristic curve (AUC) of 0.939, even with simulated metagenomic scaffolds.
• The classifier demonstrated effectiveness on both complete and partial genome sequences.

Conclusions:

• SourceFinder accurately predicts the source of bacterial sequences, aiding in the study of mobile genetic elements.
• Accurate identification of extra-chromosomal elements is crucial for understanding gene spread, antimicrobial resistance, and virulence.
• The SourceFinder web service is available to the scientific community for analyzing assembled bacterial sequence data.