PlasForest

PlasForest classifies contigs and scaffolds as plasmid or chromosomal sequences by combining homology searches against a plasmid sequence database with a random forest classifier to detect plasmids in unassembled genomic datasets.

Key Features:

• Homology-Based Search: Leverages homologies against a curated database of plasmid sequences to identify plasmid-derived contigs and scaffolds.
• Random Forest Classifier: Uses a random forest algorithm to discriminate plasmid from chromosomal sequences.
• Performance Metrics: Achieves 96% sensitivity for plasmid contigs >50 kb, a 3.3% false positive rate, and 98% overall accuracy.
• Taxonomy-Independent Classification: Operates without requiring prior taxonomic information.
• Applicability to Unassembled Data: Targets detection in unassembled genomic datasets (contigs and scaffolds), improving plasmid identification in incomplete assemblies.
• Comparative Performance: Demonstrates higher sensitivity and precision compared with existing bioinformatics methods on unassembled datasets.

Scientific Applications:

• Bacterial Genomics: Identification of plasmids to study bacterial genome composition and accessory gene content.
• Antibiotic Resistance Surveillance: Tracking dissemination of antibiotic resistance genes carried on plasmids.
• Horizontal Gene Transfer Studies: Investigating plasmid-mediated horizontal gene transfer and bacterial adaptation.
• Analysis of Unassembled Sequencing Data: Detection of plasmids in datasets where complete assembly is not feasible (contigs and scaffolds).

Methodology:

Determine homologies between genomic contigs/scaffolds and a curated plasmid sequence database and input homology-derived features into a random forest classifier to distinguish plasmid-origin contigs from chromosomal ones.