Identification and characterisation of bacterial pathogens through Large Language Model-assisted text mining

Large Language Models identify 1,222 bacterial species linked to human infection, with 783 confirmed as pathogens based on multiple literature reports.

• Why it matters: Understanding bacterial pathogen diversity is vital for addressing infection risks amid rising antimicrobial resistance and changing global conditions.
• What they did: A scalable, automated pipeline used LLMs to analyze tens of thousands of PubMed abstracts, extracting data on pathogenicity, ecological traits, and pathogen classification.
• The result: This approach creates an open, evidence-based catalogue of bacterial pathogens, enabling improved public health surveillance, diagnostics, and predictive modeling.