🦠

Overview of CARD and AMR

Oct 4, 2025

Overview

This lecture introduces the Comprehensive Antibiotic Resistance Database (CARD), its ontology (ARO), associated curation practices, and core software platforms for detecting antimicrobial resistance (AMR) genes in pathogens.

Antimicrobial Resistance & CARD

  • Antimicrobial resistance (AMR) is a global health crisis, worsened by the COVID-19 pandemic.
  • CARD is a curated knowledge base on the molecular and genetic basis of AMR.
  • CARD integrates the Antibiotic Resistance Ontology (ARO) within the OBO Foundry.
  • As of this month, CARD contains nearly 7,000 ontology terms, 5,000+ reference sequences, and data from 3,000+ publications.

CARD Structure & Ontology

  • The ontology is divided into seven branches; four core: AMR determinants (genes), mechanisms, antibiotic molecules, and targets.
  • Ontology connects terms with defined semantic relationships and is regularly reviewed and updated.

AMR Detection Models & Curation

  • Each AMR determinant in CARD needs a peer-reviewed publication, a public sequence, and experimental evidence of resistance.
  • Detection models in CARD include DNA/protein sequences, resistance-associated mutations, and a gene-specific bit score cutoff reflecting detection strictness.
  • Community feedback and targeted literature reviews guide curation.
  • The CardShark machine learning tool prioritizes literature for curators.

Resistance Gene Identifier (RGI) Software

  • RGI software allows users to identify resistance genes in isolates by comparing to CARD’s models.
  • RGI reports “perfect” (100% match) and “strict” (above bit score cutoff) hits.
  • RGI is available online and via command line; a version, RGI BWT, is optimized for metagenomic samples.

CARD Resistomes & Variants Project

  • CARD now includes large-scale AMR prediction across over 377 pathogens and 200,000 genome assemblies.
  • Outputs gene-level and drug class-level resistance predictions (perfect/strict hits) for each isolate.
  • Data are online and downloadable, supporting pathogen- and gene-centric queries.

Ongoing Challenges & Developments

  • Enhancing RGI with better annotation and context awareness for more accurate predictions.
  • Considering machine learning for non-genotypic resistance (e.g., biofilms).
  • Exploring a shift from gene-centric to pathogen-centric data views.
  • Initiating a nomenclature standardization project for aminoglycoside modifying enzymes.

Key Terms & Definitions

  • Antimicrobial Resistance (AMR) — The ability of microbes to survive treatments with drugs designed to kill them.
  • CARD — Comprehensive Antibiotic Resistance Database; a resource of AMR gene information.
  • ARO — Antibiotic Resistance Ontology; a structured vocabulary for AMR concepts within CARD.
  • Detection model — A CARD entry consisting of a gene sequence, relevant mutations, and a bit score cutoff for AMR identification.
  • Bit score cutoff — A statistical threshold determining the similarity required for a gene to be considered a match in CARD.
  • RGI — Resistance Gene Identifier; software for detecting AMR genes using CARD.

Action Items / Next Steps

  • Explore CARD and ARO resources online for further learning.
  • Test RGI or RGI BWT with your own isolate sequence data if interested in AMR analysis.
  • For those interested in nomenclature standardization, contact the CARD team.
  • Consider reviewing new publications or giving feedback to assist CARD curation.

Full transcript