AOPTK: Open-Source Software for AI-Assisted Toxicological Literature Analysis

AOPTK, an open-source Python toolkit for AI-assisted toxicology literature analysis, achieves 81% agreement with an expert inhibitor list and 95% accuracy on 86 publications.

• Why it matters: Toxicologists struggle to extract and interpret evidence across thousands of studies, limiting use of available chemical–health information. This work targets scalable automation for toxicological literature review.
• What they did: The toolkit automates literature retrieval from PubMed Central and Europe PMC, processes PDFs, and uses LLM-assisted tools to identify chemicals and extract relationships to biological effects under user-defined conditions. It was evaluated on a subset of 86 publications for predicting inhibition versus no inhibition.
• The result: In that subset, LLM-enhanced analysis matched expert curation with 81% agreement and reached 95% accuracy for inhibition prediction. It also analyzed 1 124 PubMed Central publications, supporting feasibility for large-scale studies.

AI summary of the abstract · may contain errors · not medical advice · read the paper.