BioLiP3: a protein-ligand affinity database integrating curated data with LLM-enabled literature mining.

BioLiP3 expands quantitative protein–ligand affinity coverage 13-fold, totaling 593,423 affinity records, with 591,820 linked to 3D structures across experimentally resolved and modeled complexes.

• Why it matters: Binding affinities and 3D protein–ligand structures are complementary for structure-based drug discovery, but quantitative data are fragmented between databases and biomedical literature.
• What they did: The team built BioLiP3 by integrating 45,386 experimentally resolved PDB complexes, 528,732 BindingDB-derived records with computational structures, and 17,702 structure-linked records from LLM-mined PubMed abstracts, extracting Ki, Kd, IC50, and EC50 with provenance.
• The result: BioLiP3 distinguishes experimental versus computational structures and provides cross-references to PDB, UniProt, NCBI Taxonomy, and PubChem, enabling unified data-driven protein–ligand interaction studies and structure-based drug design.

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