Open-Source Tool Uses AI to Mine Toxicology Research

The AOPTK Python package applies large language models to extract chemical and biological relationships from published studies.

Updated on Nov. 9, 2026 in Artificial Intelligence

Open-Source Tool Uses AI to Mine Toxicology Research

Researchers have released AOPTK, an open-source Python package that uses large language models to analyze toxicological and biomedical literature. In a gap junction case study, the tool's LLM-enhanced analysis reached 81% agreement with an expert-curated inhibitor list.

Why it matters

Much toxicological evidence goes underutilized because manually extracting and interpreting findings across thousands of studies is impractical. Automated literature analysis could help researchers surface relevant data more efficiently.

In a case study, the LLM-enhanced analysis matched an expert-curated inhibitor list 81% of the time and predicted inhibition versus no inhibition with 95% accuracy across 86 publications. The tool's performance beyond this gap junction case study has not been reported.

The players

AOPTK

An open-source Python package for AI-enhanced analysis of toxicological and biomedical literature, developed on GitHub and distributed through PyPI and Bioconda.

The details

AOPTK is an open-source Python package that retrieves articles automatically from PubMed Central and Europe PMC and processes full-text PDFs directly. Its LLM-assisted tools identify chemicals and extract their relationships to biological effects under user-defined conditions, and the case study used this approach to flag potential non-genotoxic carcinogens based on their ability to inhibit gap junction intercellular communication.

Timeline

  1. November 9, 2026: Article describing the AOPTK release is published.

Roadmap

AOPTK applies large language models to the long-running problem of turning thousands of published toxicology studies into usable evidence. It extends efforts to automate biomedical literature review into a domain where manual extraction has been especially slow.

Researchers working on chemical safety and toxicology may be able to surface relevant findings from large volumes of published studies with less manual review. The tools are available free as open-source software, though their adoption across the field has not been reported.

The takeaway

Automated tools that combine literature retrieval with language-model analysis could make underused toxicological evidence easier to access. Their reliability outside a single case study will determine how widely they are adopted.

Further reading

For related coverage, see Artificial Intelligence on Wisevoter.

Source note: This article includes information reported by Biorxiv.