spatial-data-validation

Spatial data validation

Validate first, reason second. Never assume a file's format or contents from its extension.

Steps

1. Detect & validate format. Call the MCP tool validate_spatial_data with the file path. Choose validation_level:

   • basic — extension/existence only.
   • structure (default) — format + structural integrity.
   • integrity — deeper content checks.
   • domain — spatial-biology-specific expectations. For several files, use validate_multiple_spatial_files.

2. Read the metadata. Call analyze_spatial_metadata to extract dimensions, spatial coordinates, and gene/feature info. Confirm obsm['spatial'] (AnnData) or coordinate systems (SpatialData) are present when spatial analysis is intended.

3. Check raw vs. processed. Integer-valued matrices are likely raw counts; non-integer/log-scaled values are processed. State which the downstream method expects and flag a mismatch.

4. Check compatibility when combining files: check_spatial_data_compatibility (coordinate systems + gene overlap).

Key facts (see context/data-formats.md)

• A .zarr store is SpatialData only if it has SpatialData markers or ≥2 of images/ labels/ points/ shapes/ tables/. Otherwise it's a plain zarr array.
• iST datasets can be terabytes. Default to lightweight structural checks; avoid loading whole objects. The heavy scientific stack (pip install -e ".[spatial]") is optional — the validators degrade gracefully without it.

Don't

• Don't claim a file is valid SpatialData without inspecting contents.
• Don't call execution tools (e.g. run_nextflow_workflow) — they aren't implemented. Drive local CLIs directly if execution is needed.