New cell type annotations now available in the results bucket

[allyhawkins]

The results from two new modules, cell-type-scimilarity and cell-type-neuroblastoma-04 along with updates to the existing cell-type-consensus module are now available in the results bucket.

• The cell-type-scimilarity module annotates all cells in all ScPCA libraries using SCimilarity.
• The cell-type-neuroblastoma-4 module annotates cell types in all libraries from SCPCP000004 using the NBAtlas reference (Bonine et al. 2024). Cells are annotated by looking for agreement between annotations from SingleR, scANVI/scArches, and the cell-type-consensus module.
• The cell-type-consensus module has been updated to now incorporate annotations from the cell-type-scimilarity module. Here, consensus cell types are assigned when annotations agree between 2 of the 3 automated methods used for cell typing (SingleR, CellAssign, and SCimilarity).

Full details on the analysis and available output for each module can be found in the module's README.md files.

To download the results from these modules, you can run the following commands:

./download-results.py --modules cell-type-scimilarity
./download-results.py --modules cell-type-neuroblastoma-04
./download-results.py --modules cell-type-consensus

Note that the above commands do require you to be logged into an AWS account with appropriate access privileges. Test data that does not require login is available by adding the --test-data flag.

For more details on options available in the download-results.py script, see the script documentation.

---

Please note that as of the 2025-06-30 data release, consensus cell type labels that are provided in the processed SingleCellExperiment and AnnData objects were assigned using only input from SingleR and CellAssign and will not match the output of the cell-type-consensus module. Future data releases will include the updated consensus cell types in the processed objects.

--

[Update 2026-02-05] The workflow results associated with the 2025-06-30 data release have been updated to fix a small bug in assignment of consensus cell types. Previously consensus cell types were unassigned if two out of three methods agreed but the third method was unable to classify the cell type. The reference file has been updated and the results have been regenerated such that this scenario correctly assigns the most appropriate cell ontology based on the two methods in agreement.