OncoMap: a curated, literature-grounded catalogue of oncology spatial-omics datasets

Spatial omics (both spatial transcriptomics and spatial proteomics) is transforming tumour biology, but the resulting public datasets are scattered across general-purpose registries (GEO, ArrayExpress), cell atlases (CZ CELLxGENE), citable archives (Zenodo), image repositories (the EBI BioImage Archive) and consortium portals (HTAN), described with inconsistent metadata and of highly variable reusability. Finding the spatial data that exists for a given cancer type, on a given platform and modality, and confirming it can actually be reused, is slow and error-prone. We present OncoMap, a curated, literature-grounded metadata catalogue that maps oncology spatial-omics datasets to their publications, methods and research groups through a versioned, ontology-linked knowledge graph of four node types and four typed edge types. Every record is validated against a JSON Schema and a three-tier curation trust gate; cancer types are coded in OncoTree (with NCI Thesaurus cross-maps) and tissues in UBERON, from frozen, membership-enforced vocabularies. The v0.1 release contains 247 datasets across 42 cancer types and 16 platforms in two modalities (175 spatial transcriptomics and 72 spatial proteomics), harvested and integrity-filtered from seven sources and verified against their source publications, and exposes a per-dataset reusability score computed from seven machine-checkable signals. OncoMap is browsable at www.oncomap.org, released under CC-BY-4.0, and built by a deterministic, tested, and monitored pipeline that makes catalogue health an observable, service-level-gated property.