Closed-form floors versus learned models in spatial omics: confirmatory protocol and registration package
Pre-registered confirmatory protocol for a benchmark of pre-specified closed-form floors against published learned models in five spatial-omics task families (cell identity, spatial domains, spatial context to receiver state, held-out cell-type counterfactuals, label transfer). Registered before any confirmation metric was computed, except the retrospective comparison with the authors' published Cellina and CPA results stated in the protocol. Contents: PROTOCOL.md (aim, hypotheses, families and comparators, data roles, analysis with margins and decision rules, power, tier-2 tuning-transfer and capacity diagnostic, deviation policy, compute, development history); spatialfloors-registration-v1.tar.gz (the full package: locked configurations, unit manifests with SHA-256 of every crop's cell-ID list, gene panels, the sealed confirmation job list, the code archive of the release commit and SHA256SUMS); SHA256SUMS. Verify: unpack, then shasum -a 256 -c SHA256SUMS inside registration/.
Authors
- Shayan Forghani (ORCID: https://orcid.org/0009-0003-4341-1299)
- Amirhossein Shahsavand (ORCID: https://orcid.org/0009-0000-9057-1680)
Institutions
- Tehran University of Medical Sciences (IR)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-03
- DOI
- https://doi.org/10.5281/zenodo.23931263
- Primary Topic
- Single-cell and spatial transcriptomics
- Type
- article
- Field-Weighted Citation Impact
- 0.00