OmniBioAI
Sign in

Single-Cell Analysis (Seurat)

Guided clustering (v1): QC → LogNormalize → HVG → PCA → Louvain → UMAP → Markers. Supports CellRanger MTX, CellRanger H5, and alevin-fry.
Plugin: single_cell_analysis Engine: local Upload limit: 4096 MB
{# ----------------------------------------- 1) Create Run (Registry or Upload) ------------------------------------------ #}

1) Create Run

Choose input type and source. The run will stage inputs under work/runs/single_cell_analysis/<run_id>/in/.
Input type
MTX/alevin-fry typically arrive as archives; H5 can be a direct .h5 or an archive containing it.
Source
Select object
Tip: If your dataset is a folder on disk, register it as a directory-like object; otherwise use an archive/H5.
Reset
After creating a run, go to the results page and click “Execute” (or wire an execute button later).
{# ----------------------------------------- 2) Upload & Register (optional helper) ------------------------------------------ #}

2) Upload & Register (optional)

Upload a dataset and (best-effort) register it into the Object Registry so it appears in the dropdown next time.
Label (optional)
File
Stored under work/uploads/single_cell_analysis/.
If you upload MTX/alevin-fry, prefer archiving the *folder* so detection works (see logs if it fails).
{# ----------------------------------------- What will be produced (v1) ------------------------------------------ #}

What v1 produces

  • out/summary.json
  • out/qc/qc_metrics_per_cell.csv, out/qc/qc_thresholds.json
  • out/pca/pca_variance.csv, out/pca/dims_selected.json
  • out/clusters/cluster_assignments.csv, out/clusters/cluster_sizes.csv
  • out/markers/markers_all_clusters.csv, out/markers/markers_top_per_cluster.csv
  • out/seurat_obj.rds
  • out/ui_manifest.json (allow-list for safe UI + read-only AI)