PlasForest: a homology-based random forest classifier for plasmid detection in genomic datasets

Area of Science:

• Genomics
• Bioinformatics
• Microbiology

Background:

• Plasmids are key mobile genetic elements facilitating horizontal gene transfer in bacteria.
• Accurate plasmid detection is vital for understanding bacterial adaptation and antibiotic resistance.
• Existing bioinformatics tools struggle with sensitivity, precision, and partially assembled genomes.

Purpose of the Study:

• To develop a sensitive and precise bioinformatics tool for plasmid detection in bacterial genomes.
• To address limitations of current methods in handling fragmented genomic data.

Main Methods:

• Development of PlasForest, a homology-based random forest classifier.
• Application to identify plasmid sequences within partially assembled bacterial genomes.

Main Results:

• PlasForest achieves a high F1 score of 0.950 for plasmid contig identification, irrespective of sample origin.
• It demonstrates high sensitivity for small (77.4% for <1 kb) and large (99.9% for >50 kb) plasmid contigs.
• False positive rates are low, at 2.8% and 2.2% for small and large contigs, respectively.

Conclusions:

• PlasForest surpasses existing tools in sensitivity and precision for plasmid detection in genomic datasets.
• Further research is needed to optimize PlasForest for metagenomic assemblies, potentially through homology-based approaches.