amica-agent
AMICA: Agentic Dataset Annotator
AMICA (Automated Mapping and Integration of Curation Annotations) is an LLM- and ontology-driven tool for automating cell type annotation in single-cell datasets, especially from HCA and CELLxGENE.
-
Github repository: https://github.com/Cellular-Semantics/amica-agent/
-
Documentation https://Cellular-Semantics.github.io/amica-agent/
🚀 What It Does
- Maps author annotations to Cell Ontology (CL) terms using:
- Ontology Lookup Service (OLS) API
- LLM-powered synonym matching
- Metadata-aware reasoning (e.g. tissue, stage)
- Flags unmatched terms as potential new ontology entries
- Suggests more specific terms using context and ontology structure
- Validates performance against curated CELLxGENE datasets