#28 · Tools & infrastructure

Expert curation anchors scalable annotation of biomedical literature and omics metadata

The common thread is curated biological knowledge as the primary product, with expert or community validation anchoring each resource. Manual curation and harmonization of metadata from metagenomic and cancer genomic repositories addresses heterogeneous, unstandardized terminology and quality problems in public omics data. SIBiLS extends this by attaching over 16 billion ontology-based annotations to about 40M MEDLINE abstracts and PMC full texts in BioC/JATS to support biocuration. GeneReL applies a large language model to extract gene regulatory relationships from literature, yielding 13,710 community-curated Arabidopsis interactions. What is new is scale: LLM-based extraction and ontology-driven annotation are being layered onto curation workflows, with human curation retained as the validation step. The metadata study addresses repository harmonization rather than AI-assisted extraction, so this shift is not shared by every item.