Elsevier Adopted MolMole AI for Chemical Data
Elsevier partnered with LG AI Research to automate the extraction of chemical structures from scientific literature.
Updated on Sept. 25, 2026 in Chemistry
Live Poll
Should AI models used for scientific research be required to be open for independent peer review?
Elsevier has launched the MolMole AI model to streamline the extraction of chemical structure data from patents and research journals. Developed with LG AI Research, the tool digitizes complex imagery to improve the curation speed of the Reaxys database.
Why it matters
The initiative aims to accelerate the processing of large volumes of chemical information that were previously manually curated. By automating this workflow, researchers can access structured substance data at a greater scale.
The system utilizes molecule detection, reaction-diagram parsing, and optical structure recognition to process image data. Every extraction is validated against existing Reaxys benchmarks before being committed to the database.
The players
Elsevier
Elsevier is a major global information and analytics company specializing in scientific, technical, and medical content.
LG AI Research
LG AI Research is the artificial intelligence division of the LG Group focused on developing advanced machine learning models.
The details
MolMole functions by scanning documents to identify and parse complex chemical diagrams that traditional text extraction often misses. While it significantly speeds up database population, the proprietary model currently faces limitations in recognizing metal-organic frameworks.
Timeline
September 25, 2026: The implementation of the AI model was formally detailed.
The Big Picture
The implementation of MolMole expands the automated input capabilities for the Reaxys database. This integration serves to upgrade the data ingestion speed for the long-standing Reaxys chemical information system.
Researchers and database users will gain faster access to high-quality chemical structure data as the Reaxys database expands more efficiently. However, users tracking metal-organic frameworks must continue relying on existing manual methods due to current system limitations.
The takeaway
Automated image parsing represents a significant shift in how scientific databases maintain accuracy at scale. Scientists should expect faster updates to structured databases as these proprietary models replace legacy manual entry processes.
Further reading
For more on how new technologies are impacting laboratory information management, see Chemistry.
Live Poll
Should AI models used for scientific research be required to be open for independent peer review?
