Skip to content

v0.7.0 — 10x Xenium in the Spatial omics mode

Choose a tag to compare

@belkassaby belkassaby released this 13 Nov 21:51
· 693 commits to main since this release

Released 2026-10-01.

10x Xenium in the Spatial omics mode: cells drawn as their segmentation outlines, every transcript at any zoom, a transcript density map and an Explorer-style controls panel, all read in place from the 10x bundle. The first host is Jax Image Tools, whose jit-service serves the same wire format.

Added

  • Tiled cells, transcripts and density on SpatialDataPort — all optional, so an existing port is unaffected:
    • getPolygonTile(set, level, gx, gy): cell and nucleus outlines at the four level-of-detail levels 10x ships.
    • getTranscriptTile(level, gx, gy, query): per-gene transcripts, or every gene with ALL_GENES, each tagged with the cell it falls in.
    • getTranscriptBins(level, tx, ty): an all-gene grouping pyramid, built ahead of time.
    • getTranscriptSummary, getTranscriptCounts and getDensity(genes, binSize) back the hover and the points estimate, and draw the density map.
    • importGroups(label, table) imports a cell_id,group table as a categorical column.
    • The matching contracts in spatial-dataset.contract, advertised on SpatialDataset as polygonTiles, transcriptTiles, transcriptBins and density.
  • napari · WebGPU rendering of those layers (NapariSpatialTileLayers):
    • Cells are drawn as outlines and/or fills, at the polygon level that suits the zoom.
    • Transcripts are drawn as circles sized by how many each stands for, or as one icon per gene, within a marker budget.
    • "All genes" switches between grouped bins and individual transcripts.
    • The density map is drawn over the cells, with its own bin size, range and colormap.
    • Hovering a cell or transcript reports its genes, cell and counts.
    • A view with a tile that failed to load is retried.
  • Explorer-style controls, in collapsible sections (Images, Cells, Transcripts, Annotations, Observations), each with an on/off switch:
    • Cells: colour by group, gene, transcript density, segmentation method or one colour, and hide groups.
    • Transcripts: a gene tree with per-gene colours and groups.
    • New SpatialViewState fields for all of this, defaulted in DEFAULT_SPATIAL_VIEW.
  • A navigator minimap for the napari image views, like OpenSeadragon's: click or drag to pan at the same zoom.
  • SpatialDataHttpService speaks the new routes. The group import goes through HttpClient, so host interceptors (auth) apply.
  • Example tile server:
    • Reads Xenium bundles in place (lib/spatial-xenium.mjs), with prepare-xenium for the tissue image and the all-gene transcript pyramid.
    • Group import is off by default: GROUP_IMPORT_TOKEN, or GROUP_IMPORT=open locally.
    • See its README and docs/design/omics-preprocessing-jit.md.

Changed

  • SpatialViewState gains required fields (the cell, transcript, density and annotation settings above). A host that builds a view state from DEFAULT_SPATIAL_VIEW is unaffected; one that writes a full literal must add them. That is a type-level break, so this is a minor bump under 0.x.

npm: @jax-data-science/sci-image-visualizer@0.7.0
Full diff: v0.6.0...v0.7.0