Elsevier Integrates LG AI’s MolMole to Automate Chemical Data Extraction for Reaxys

Elsevier Deploys LG AI Research Tool Inside Reaxys

The collaboration seeks to drastically shorten data capture timelines for chemists.

Automating Molecular Extraction with MolMole

The system unites molecule detection, reaction-diagram parsing, and optical chemical structure recognition inside a unified model designated as MolMole.

Elsevier reports that the architecture processes substance details from journal and patent imagery at a velocity and volume far surpassing previous possibilities.

“Every hour a chemist spends deciphering figures or images to see what has already been made is an hour that could instead be spent on chemistry discovery,” says Mirit Eldor, managing director, life sciences, at Elsevier.

Validation Protocols and Technical Boundaries

Every extracted data point passes through a verification check against established Reaxys benchmarks prior to deployment. Lutz Weber, co-founder of German software firm MolGenie, argues that utilizing these derivative methods delivers steep productivity increases over entirely manual curation workflows.

Even so, technical hurdles persist. Weber points out that the software remains incapable of extracting metal-organic complexes or frameworks.

Independent Researchers Cry Foul on Closed-Source Models

Weber observes that LG’s method stays walled off from independent testing because its source code is withheld.

With no public interface available for image-to-structure conversion, outside experts cannot benchmark the tool against alternatives like Decimer or MolScribe. Compounding the issue, the foundational research paper debuted on arXiv absent formal peer review.

Community Verification Blocked by Commercial Restrictions

Christoph Steinbeck, an analytical chemist at Friedrich-Schiller-University Jena whose team builds Decimer, shared sharp critiques of the rollout.

“Plausible numbers, a genuinely valuable idea and no way for an independent party to check either of them,” Steinbeck says.

Steinbeck emphasizes that strict licensing terms prohibit commercial use and derivative works. That legal barrier effectively chokes off industry benchmarking and thwarts community-driven verification efforts.