Skip to main content
Ctrl+K

Nicheverse

  • Gallery
  • API reference
  • Installation
    • Viewer
    • Niche universe
    • 01 Quickstart: train a Nicheverse codebook
    • 02 Transcript context: the segmentation-free molecular field
    • 03 Molecule set: the subcellular transcript point cloud
    • 04 Apply a trained model to new data
    • 05 Annotate codes into cell types and spatial niches
    • 06 Downstream analysis: using the codes and embeddings
    • Nicheverse on 10x Xenium
    • Nicheverse on Vizgen MERFISH
    • Quickstart
    • Inputs and outputs
    • Choosing hyperparameters
    • Annotating codes
    • MCP server (Claude Code / Codex tools)
  • GitHub
  • Gallery
  • API reference
  • Installation
  • Viewer
  • Niche universe
  • 01 Quickstart: train a Nicheverse codebook
  • 02 Transcript context: the segmentation-free molecular field
  • 03 Molecule set: the subcellular transcript point cloud
  • 04 Apply a trained model to new data
  • 05 Annotate codes into cell types and spatial niches
  • 06 Downstream analysis: using the codes and embeddings
  • Nicheverse on 10x Xenium
  • Nicheverse on Vizgen MERFISH
  • Quickstart
  • Inputs and outputs
  • Choosing hyperparameters
  • Annotating codes
  • MCP server (Claude Code / Codex tools)
  • GitHub
  • Gallery

Gallery#

Explore the atlases mapped in the nicheverse

One frozen model, read across every tissue.

… results

Each map is produced by loading the released checkpoint, running nicheverse.predict_codes on the dataset, annotating every cell-state code with a literature-grounded cell type, and coloring each cell by the lineage of its code. The catalogue and the full-resolution vector maps regenerate automatically as datasets are added.