You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
PR #64 updated Space Ranger to 4.1.0 and exposed the new 4.x arguments for nucleus/cell segmentation (--nucleus_segmentation, --custom_segmentation_file, --nucleus_expansion_distance_micron, --max_nucleus_diameter_px), UMI-based registration (--umi_registration, --umi_to_image_offset), cell-type annotation (--cell_annotation_model, --tenx_cloud_token_path, --disable_cell_annotation), and --include_introns.
These were reviewed and merged with unit-level coverage (optional --probe_set, --unknown-slide no-slidefile path, converter without a probe set). However, the segmentation/annotation features only trigger on Visium HD / HD 3' data with an H&E image, so they can't be exercised by the current FFPE (visium) fixture.
Why deferred
Truly testing these end-to-end requires:
Running Space Ranger 4.1 on the Visium HD fixture with an H&E image so segmented_outputs/ (+ nucleus_segmentations.geojson, cell_types/) are produced.
An HD-aware from_spaceranger_to_h5mu (or a new converter): the current one reads a flat bundle and CSV tissue_positions, whereas HD output uses binned_outputs/square_XXXum/, segmented_outputs/, and parquet tissue positions.
An HD ingestion workflow in openpipeline_spatial (currently HD ingestion lives in pfizer-spatial).
Scope of this issue
HD-aware output handling in the converter (bins vs cells, parquet positions, segmented_outputs/).
An HD (or HD 3') integration test that runs spaceranger_count with segmentation enabled and asserts the segmentation/annotation outputs.
Coverage for --include_introns and the segmentation/annotation params on real HD data.
Also relevant: resources_test_scripts/visium_hd_tiny.sh currently downloads a probe set for an HD 3' dataset, but per 10x, Visium HD 3' must NOT be given --probe_set -- worth reconciling as part of this work.
Context
PR #64 updated Space Ranger to 4.1.0 and exposed the new 4.x arguments for nucleus/cell segmentation (
--nucleus_segmentation,--custom_segmentation_file,--nucleus_expansion_distance_micron,--max_nucleus_diameter_px), UMI-based registration (--umi_registration,--umi_to_image_offset), cell-type annotation (--cell_annotation_model,--tenx_cloud_token_path,--disable_cell_annotation), and--include_introns.These were reviewed and merged with unit-level coverage (optional
--probe_set,--unknown-slideno-slidefile path, converter without a probe set). However, the segmentation/annotation features only trigger on Visium HD / HD 3' data with an H&E image, so they can't be exercised by the current FFPE (visium) fixture.Why deferred
Truly testing these end-to-end requires:
segmented_outputs/(+nucleus_segmentations.geojson,cell_types/) are produced.from_spaceranger_to_h5mu(or a new converter): the current one reads a flat bundle and CSVtissue_positions, whereas HD output usesbinned_outputs/square_XXXum/,segmented_outputs/, and parquet tissue positions.Scope of this issue
segmented_outputs/).spaceranger_countwith segmentation enabled and asserts the segmentation/annotation outputs.--include_intronsand the segmentation/annotation params on real HD data.Related
resources_test_scripts/visium_hd_tiny.shcurrently downloads a probe set for an HD 3' dataset, but per 10x, Visium HD 3' must NOT be given--probe_set-- worth reconciling as part of this work.