Data
The Data section prepares everything downstream analysis needs.
Datasets
Upload or convert data files:
| Format | Notes |
|---|---|
.h5ad | Preferred; validated at ingest |
| CSV/TSV count matrix | Convert to AnnData via Convert |
10x HDF5 (.h5) | Convert to AnnData |
| 10x MTX zip bundle | Convert to AnnData |
After upload, the dataset row shows cell/spot counts, gene count, and ingestion status. Failed ingestions display an error message you can use to fix the file.
Metadata
Define a metadata table keyed to barcodes or sample IDs in the dataset. Columns typically include sample, condition, donor, and batch. Before each analysis run the platform validates that join keys match and merges columns into AnnData obs.
Use Validate against dataset to catch mismatched IDs before running expensive jobs.
Design
Specify which obs columns define biological groups and which column identifies biological replicates. The design table feeds contrast definitions and future bulk statistical models.
Contrasts
Save named comparisons (for example Treatment vs Control on the condition column). Contrasts appear in the Analyze DE step when comparison mode is condition or sample.
Ingestion
The Ingestion panel summarizes format detection, conversion jobs, and any sidecar files used during CSV/10x import.
Tips
- Run metadata validation after every metadata edit; stale flags appear on Runs if you change metadata after DE.
- Spatial datasets must include
obsm['spatial']coordinates in the.h5ad. - Biomarker workflows need a label column with at least two classes (default:
conditioninobs).