CARDShark

CARDShark automates prioritization of PubMed publications to identify first reports of new antimicrobial resistance genes for curation into the Comprehensive Antibiotic Resistance Database (CARD).

Key Features:

• Automated Prioritization: A classification algorithm trained on existing CARD publications ranks newly published articles likely to contain novel resistance gene reports.
• Machine Learning Classification: Uses a machine learning-based classifier trained on the CARD dataset to recognize patterns indicative of first reports of antimicrobial resistance genes.
• PubMed Processing and Triage: Continuously downloads and processes recent additions to PubMed to filter out irrelevant publications and flag pertinent studies.
• Reduced Curation Scope: Narrows the monthly set of articles requiring manual review from hundreds to a few dozen, focusing biocurator effort on high-priority publications.

Scientific Applications:

• CARD Curation: Accelerates the identification and integration of newly reported antimicrobial resistance genes into the Comprehensive Antibiotic Resistance Database (CARD).
• Resistance Surveillance: Supports tracking of emerging antimicrobial resistance genes and trends across the scientific literature.
• Public Health and Policy Support: Provides timely evidence to inform research, clinical responses, and public health decision-making related to antibiotic resistance.

Methodology:

Applies a machine learning-based classification algorithm trained on the CARD dataset and continuously downloads and processes new PubMed publications, applying the learned model to identify and flag relevant articles for biocurator review.