ROCker

ROCker models facilitate reliable detection and typing of short-read sequences carrying antibiotic resistance genes in unassembled metagenomic, metatranscriptomic, and PCR amplicon data.

Key Features:

• Target Gene Coverage: ROCker provides models for mcr, erm, mph, and lnu ARG families, including variants of special public health concern such as mcr-1, mphA, ermB, lnuF, lnuB, and lnuG.
• Improved Classification Performance: ROCker models achieve F1 scores up to 10x higher (31% higher on average) compared to alternative methods including BLASTx with various e-value thresholds and hidden Markov models.
• Reduced Error Rates: ROCker models demonstrate lower false-positive rates (by 30% on average) and false-negative rates (by 16% on average) based on 250 bp reads compared to static filtering approaches.
• Model Availability: All ROCker models, reference materials, and training data are freely accessible through the enve-omics website.

Scientific Applications:

• Antimicrobial Resistance Surveillance: ROCker enables quantitative monitoring of emerging ARGs in clinical and environmental samples to assess prevalence and track spread to new pathogens.
• Short-Read Metagenomics: ROCker facilitates accurate classification and quantification of ARG-carrying sequences in unassembled short-read metagenomes.
• Metatranscriptomics: ROCker supports detection and typing of ARG sequences in metatranscriptomic data.
• PCR Amplicon Analysis: ROCker can be applied to PCR amplicon data for ARG detection and variant identification.

Methodology:

ROCker models are built using the ROCker V2 pipeline, which involves curation of target gene sequence sets for model training. Models are validated using simulated reads generated from whole genomes of ARG-carrying isolates across a range of common read lengths.