Skip to content

SpatialData input with points + shapes fails to use prior segmentation (--zarr-shape) in v3.0.12 #133

@kimdansay

Description

@kimdansay

I am trying to run ProSeg v3.0.12 using a SpatialData zarr input that contains:

  • transcript coordinates as points
  • cell boundary geometries as shapes
  • intended use of prior cell segmentation via --zarr-shape

However, ProSeg fails during input parsing and does not correctly handle this SpatialData structure, even though it follows the SpatialData specification.

[Environment]

  • ProSeg version: 3.0.12
  • Installation: cargo install proseg
  • OS: Linux (HPC cluster)
  • SpatialData generated with: spatialdata (Python)

[SpatialData input structure]

The input zarr has the following structure:
xenium_proseg_input.zarr
├── points/
│ └── transcripts/
│ └── points.parquet
├── shapes/
│ └── cell_boundaries/
├── zmetadata

Details:

  • points/transcripts contains transcript-level coordinates (x, y)
  • shapes/cell_boundaries contains polygon geometries for cells
  • The shapes are intended to be used as prior segmentation

Command used
proseg
--zarr xenium_proseg_input.zarr
--zarr-shape xenium_proseg_input.zarr
--zarr-shape-geometry-column geometry
--zarr-shape-cell-id-column cell_id
--use-cell-initialization
--prior-seg-reassignment-prob 0.3

ProSeg terminates with the following error:
Missing required argument: --x
note: run with RUST_BACKTRACE=1 environment variable to display a backtrace

This occurs even though transcript coordinates are present in points/transcripts and no table-based or CSV-based input is being used.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions