LEADO: unlocking functional insights from lipidomics data through mining lipid-gene-disease relationships.

With the rapidly expanding repertoire of literature in the life sciences, efficiently extracting and integrating critical biomedical information from texts has emerged as an urgent challenge. Fuelled by the development of lipidomics, lipids have garnered increasing attention for their roles as essential molecules that regulate cellular signalling and metabolism, which underlie the pathogenesis of various diseases. Existing databases, however, primarily focus on genes, proteins, or small metabolite functions, with limited coverage of lipid-disease relationships, confining lipidomics application to a correlative framework. To address this gap, we developed Lipid Enrichment Analysis of Disease Ontology (LEADO), a BERT-based literature-mined knowledge database (https://lipidall-leado.com/leado/). We fine-tuned the BioBERT model to extract lipid entities and infer lipid-gene-disease associations, then constructed an interactive web platform using Shiny for data visualization and querying. LEADO serves as a comprehensive and user-friendly resource for exploring lipid-gene-disease relationships, with the potential to transform lipidomics toward a functional, causality-driven discipline that accelerates translational discovery in lipid metabolism.