Skip to content

Supported Platforms & Inputs

run_together's one-command convenience comes from platform profiles that read each platform's standard output layout automatically. This page is the status hub; each built-in profile has its own page with the exact expected inputs and example commands.

For the profile mechanics (input modes, the canonical config, layer merging), see Specifying Inputs.

Profile status

--platform Profile Ingest Details
10x_xenium ✅ built-in sge_convert (10x_xenium) Xenium page
10x_visium_hd ✅ built-in sge_convert (10x_visium_hd) Visium HD page
cosmx_smi ✅ built-in reformat_cosmx CosMx SMI page
merfish (Vizgen MERSCOPE) ✅ built-in sge_convert (vizgen_merscope) MERSCOPE page
illumina (Illumina StrataMap) ✅ built-in sge_convert (illumina) Illumina StrataMap page
stereoseq (BGI Stereo-seq) ✅ built-in SAW gef2gem + sge_convert (bgi_stereoseq) Stereo-seq page
seqscope (Seq-Scope) ✅ built-in sge_convert (seqscope) Seq-Scope page
generic (any source) ✅ built-in an already-ingested transcript TSV, or sge_convert (generic) Generic page
Pixel-seq, Nova-ST via a custom profile sge_convert (pixel_seq, nova_st) below
  • built-in — run_together --platform <name> works with no extra configuration. The profiles are assets/run_together_profiles/<name>.json.
  • via a custom profile — sge_convert already has an ingest preset; there is no dedicated run_together profile yet.

Built-in platforms

Each page documents the expected input files/columns, the FICTURE/packaging defaults, and example commands for the input modes (single-sample CLI, sample sheet, config JSON):

  • 10x Xenium — Xenium Ranger directory; morphology images auto-detected; xeniumranger clusters.
  • 10x Visium HD — Space Ranger outs/; square-bin and segmented-cell layers; a cartloader cell analysis on the segmented cells; per-sample H&E.
  • CosMx SMI — AtoMx flat-file export via reformat_cosmx; glob-based file matching with overrides.
  • MERSCOPE / MERFISH — individual files or a standard export dir; cell analysis from boundaries, a cell×gene matrix, or an existing cell_id (incl. mixed joint runs).
  • Illumina StrataMap — MEX input with spatial coordinates embedded in the barcode; optional cell boundaries enable cell-level analysis.
  • BGI Stereo-seq — prefix-addressed binary GEFs expanded by SAW (--saw); cell-bin segmentation clustered from its own pixel TSV; registered histology at 0.5 µm/pixel.
  • Seq-Scope — --in-dir is the MEX directory (or a raw per-molecule TSV via --raw-transcript); nanometer coordinates; pixel-level only (no segmentation); pre-georeferenced H&E.
  • Generic — a transcript TSV from any source (already in CartLoader's layout, or a CSV/TSV converted by sge_convert --platform generic); optional cell analysis from a cell-id column or a MEX.

Custom profiles

--platform-json <file> deep-merges a JSON file over the built-in profile named by --platform: nested blocks merge key by key, while lists (e.g. ingest.extra_flags, ingest.autodetect, images) are replaced whole. --platform must still name a built-in profile, so a custom profile is always a modification of one — start from the closest match, or from generic for a plain transcript table. The file is not checked for unknown keys.

For example, a Pixel-seq transcript table through sge_convert's pixel_seq preset:

1
{ "ingest": { "sge_platform": "pixel_seq" } }

1
2
cartloader run_together --platform generic --platform-json pixelseq.json \
    --raw-transcript /data/pixelseq/transcripts.tsv.gz --out-dir OUT

ingest.sge_platform selects the sge_convert preset (column names, delimiter); add --units-per-um if the coordinates are not in microns. Any profile key can be set this way; a --config can carry the same blocks for a single run (see Specifying Inputs → Mode 3), with strict key validation.

Without run_together, the modules can also be run directly: sge_convert → run_ficture2 → run_cartload2 (see the platform tutorials and sge_convert for the per-platform ingest presets).

See also