VAMPr

VAMPr predicts antimicrobial resistance (AMR) phenotypes and maps gene-level sequence variants to identify genetic determinants of resistance from whole genome sequencing data.

Key Features:

• Gene Ortholog-Based Sequence Feature Derivation: Derives gene ortholog-based sequence features for protein variants to enable detailed examination of genetic variations linked to antibiotic resistance.
• Interrogation of Gene-Level Variants: Systematically interrogates explainable gene-level variants to identify known and novel associations with AMR.
• Model Building for AMR Prediction: Constructs prediction models from whole genome sequencing data using curated sequences from 3,393 bacterial isolates across nine species with AMR phenotypes for 29 antibiotics.
• High Accuracy and Validation: Achieves a mean prediction accuracy of 91.1% across antibiotic–pathogen combinations, validated by internal nested cross-validation and external clinical datasets.
• Association Models: Produces association models that confirm known genetic resistance mechanisms, such as blaKPC linked to carbapenem resistance.

Scientific Applications:

• Clinical Utility: Predicts AMR phenotypes from genomic data to aid diagnosis of resistant infections.
• Research Advancements: Identifies previously unrecognized AMR mechanisms and supports basic research into resistance pathways.
• Data Integration: Integrates publicly available sequencing data with phenotypic information as a resource for bacterial genomics and antibiotic resistance studies.

Methodology:

Detects variant genotypes from genomic sequences and derives gene ortholog-based sequence features for protein variants; builds association models linking genetic variations to known AMR mechanisms and constructs prediction models to forecast AMR phenotypes from whole genome sequencing data, with validation by nested cross-validation and external clinical datasets using curated sequencing data from 3,393 isolates across nine species covering 29 antibiotics.