Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
BioMiner is introduced as a multi-modal framework for automated extraction of protein-ligand bioactivity data from scientific literature, separating bioactivity semantic interpretation from chemical structure construction. It uses a chemical-structure-grounded visual semantic reasoning (CSG-VSR) paradigm, leveraging multi-modal LLMs on chemically grounded visual representations. Evaluated on the newly created BioVista benchmark, BioMiner achieves an F1 score of 0.33 for complex bioactivity triplets and demonstrates practical utility in building pre-training databases, improving QSAR models, and accelerating annotation tasks.
Automating bioactivity data extraction from papers is now possible, with BioMiner achieving a 5x speedup and 10% accuracy boost over manual methods.
Protein-ligand bioactivity data published in literature are essential for drug discovery, yet manual curation struggles to keep pace with rapidly growing literature. Automated bioactivity extraction remains challenging because it requires not only interpreting biochemical semantics distributed across text, tables, and figures, but also reconstructing chemically exact ligand structures (e.g. Markush structures). To address this bottleneck, we introduce BioMiner, a multi-modal extraction framework that explicitly separates bioactivity semantic interpretation from ligand structure construction. Within BioMiner, bioactivity semantics are inferred through direct reasoning, while chemical structures are resolved via a chemical-structure-grounded visual semantic reasoning (CSG-VSR) paradigm, in which multi-modal large language models operate on chemically grounded visual representations to infer inter-structure relationships, and exact molecular construction is delegated to domain chemistry tools. For rigorous evaluation and method development, we further establish BioVista, a comprehensive benchmark comprising 16,457 bioactivity entries curated from 500 publications. On BioVista, BioMiner validates its extraction ability and provides a quantitative baseline, achieving F1 scores of 0.33 for complex bioactivity triplets. BioMiner’s practical utility is demonstrated via three applications: (1) extracting 82,262 data from 11,683 papers to build a pre-training database that improves downstream models performance by 3.9%; (2) enabling a human-in-the-loop workflow that doubles the number of high-quality NLRP3 bioactivity data, helping 38.6% improvement over 28 QSAR models and identification of 16 hit candidates with novel scaffolds; and (3) accelerating the annotation of the protein-ligand structures in PoseBusters benchmark with reported bioactivity, achieving a 5-fold speed increase and 10% accuracy improvement over manual methods. BioMiner and BioVista provide a scalable extraction methodology and a rigorous benchmark, paving the way to unlock bioactivity data that previously required extensive human effort. All codes and data are available at GitHub.