Plasmer

Plasmer predicts bacterial plasmid sequences within genomic assemblies to identify plasmids involved in horizontal gene transfer, antibiotic resistance dissemination, host–microbe interactions, and biotechnological applications.

Key Features:

• Machine learning: Employs a random forest algorithm to classify contigs as plasmid or chromosomal sequences.
• Shared k-mer analysis: Analyzes the percentage of shared k-mers with known plasmid and chromosome databases to inform classification.
• Genomic feature integration: Incorporates genomic features including alignment E values and replicon distribution scores (RDS) alongside k-mer features.
• Species-agnostic model: Functions across multiple bacterial species without requiring species-specific models.
• High performance metrics: Reports an average area under the curve (AUC) of 0.996 and an accuracy of 98.4% across benchmark tests.
• Balanced sensitivity and specificity: Achieves sensitivity and specificity exceeding 0.95 for contigs longer than 500 base pairs (bp) and attains the highest F1-score among compared methods.
• Fragmented assembly capability: Maintains high accuracy on fragmented short-read assemblies, including contigs as short as 500 bp.
• Taxonomy classification: Provides taxonomy classification features to assist in identifying the origin of detected plasmids.

Scientific Applications:

• Horizontal gene transfer studies: Detects plasmids to support analyses of gene transfer between bacteria.
• Antibiotic resistance surveillance: Identifies plasmid-borne resistance determinants for studies of resistance dissemination.
• Microbial ecology and host–microbe interactions: Enables detection of plasmids relevant to ecological studies and host-associated microbiomes.
• Cloning vectors and biotechnology: Supports identification of plasmid sequences relevant to cloning vector characterization and industrial biotechnology applications.
• Next-generation sequencing analyses: Applied to fragmented assemblies and short contigs typical of NGS datasets for plasmid discovery.

Methodology:

Integrates shared k-mer analysis (percentage of shared k-mers against known plasmid and chromosome databases) with genomic features including alignment E values and replicon distribution scores (RDS), and trains a random forest classifier; performance evaluated on sliding sequences, simulated contigs, and de novo assemblies.