Automating scientific annotations for open transcriptomic profiles via multi-stage agents

Public transcriptomic repositories contain millions of samples, yet their large-scale reuse is hindered by heterogeneous and inconsistently reported metadata. In the Gene Expression Omnibus (GEO), key biological information is often distributed across study- and sample-level records, requiring context-dependent interpretation. Here we present GEOMeta, a large language model (LLM)-based multi-stage workflow with task-specialized agents for automated GEO metadata curation. The pipeline separates metadata retrieval, task-specific information extraction, field standardization, ontology mapping and quality control. Using GEOMeta, we generated standardized annotations for approximately 600,000 human bulk RNA-seq samples. To demonstrate its utility, we benchmarked transcriptome representation models for predicting sex, age, tissue and disease from transcriptome embeddings. We further prospectively annotated newly submitted GEO studies and evaluated 22 frontier LLMs. Recent open-source Flash models achieved annotation quality comparable to leading reasoning models while reducing costs by an order of magnitude. GEOMeta provides a scalable resource and reproducible framework for metadata curation.