Researchers have developed BioMiner, a novel multi-modal system designed to automatically extract protein-ligand bioactivity data from scientific literature. This system addresses the challenge of manual curation by interpreting semantic information from text, tables, and figures, and reconstructing complex chemical structures. BioMiner achieves an F1 score of 0.32 for bioactivity triplets and has demonstrated practical utility in building pre-training databases, improving quantitative structure-activity relationship (QSAR) models, and accelerating bioactivity annotation. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Automates extraction of critical bioactivity data from literature, accelerating drug discovery and improving downstream model performance.
RANK_REASON This is a research paper describing a new system and benchmark for automated data extraction from scientific literature.
