RFPlasmid

RFPlasmid predicts plasmid-derived contigs from short-read assembly data to distinguish plasmid and chromosomal sequences for investigation of molecular epidemiology and antimicrobial resistance (AMR) gene localization.

Key Features:

• Machine learning classification: Applies machine learning to classify contigs from short-read assemblies as plasmid or chromosomal.
• Multi-feature integration: Combines multiple sequence-derived features to improve discrimination between plasmid and chromosomal contigs.
• k-mer composition analysis: Uses k-mer composition of contigs as a predictive feature.
• Marker protein databases: Queries databases of plasmid- and chromosome-specific marker proteins to inform classification.
• Species-specific models: Includes trained models for 17 bacterial species, including Campylobacter, E. coli, and Salmonella.
• Species-agnostic model: Provides a species-agnostic model suitable for metagenomic assemblies or organisms without species-specific models.
• Custom model training: Supports training custom models using user-supplied contigs labeled as chromosomal or plasmid.

Scientific Applications:

• AMR gene localization: Determine whether antimicrobial resistance (AMR) genes are located on plasmids or chromosomes to assess potential horizontal transfer.
• Molecular epidemiology: Support studies of plasmid-mediated spread and molecular epidemiology of bacterial pathogens.
• Metagenomic assembly analysis: Classify contigs in metagenomic assemblies using the species-agnostic model.
• Plasmid identification in draft genomes: Distinguish plasmid from chromosomal contigs in fragmented draft whole-genome assemblies, including cases with large single-copy plasmids.

Methodology:

Classifies contigs from short-read assemblies using machine learning with features including k-mer composition and matches to plasmid- and chromosome-specific marker protein databases, using species-specific and species-agnostic models and allowing custom model training on labeled contig datasets.