PlasmidHostFinder

PlasmidHostFinder predicts the host range of plasmids to identify potential bacterial hosts and support investigations of plasmid-mediated antimicrobial resistance dissemination.

Key Features:

• Machine Learning Approach: Uses a random forest algorithm to detect complex sequence-based patterns for host prediction.
• Comprehensive Training Dataset: Trained on a dataset of 8,519 plasmids from 359 bacterial species across multiple taxonomic levels.
• High Predictive Accuracy: Reports Matthews correlation coefficients of 0.662 at the species level and 0.867 at the order level.
• Pattern-detection vs Homology: Employs a pattern-detection approach that is reported to outperform traditional homology-based methods given plasmid genetic diversity and plasticity.

Scientific Applications:

• Surveillance of Antimicrobial Resistance: Predicts potential hosts to track dissemination pathways of antimicrobial resistance genes among bacteria.
• Bacterial Evolution Studies: Informs analyses of horizontal gene transfer and evolutionary trajectories of plasmid-bearing bacteria.
• Public Health Interventions: Identifies candidate host organisms to inform strategies aimed at mitigating plasmid-mediated spread of resistance.

Methodology:

Implements a random forest model trained on 8,519 plasmids from 359 bacterial species and applies a pattern-detection approach rather than homology-based methods.