umap.json

{
  "tool_id": "umap",
  "name": "UMAP Embedding",
  "category": "dimensionality_reduction",
  "framework": "scanpy",
  "function": "sc.tl.umap",
  "valid_after": ["neighbor_graph"],
  "valid_before": [],
  "paradigms": ["all"],
  "parameters": {
    "min_dist": {
      "type": "float",
      "default": 0.5,
      "range": [0.0, 1.0],
      "guidance": "Controls how tightly UMAP packs points together. Lower values produce tighter clusters visually. 0.5 is default; 0.1-0.3 often produces more visually distinct clusters."
    },
    "spread": {
      "type": "float",
      "default": 1.0,
      "range": [0.5, 3.0],
      "guidance": "Scale of the embedding. Usually left at 1.0."
    },
    "n_components": {
      "type": "int",
      "default": 2,
      "guidance": "Number of UMAP dimensions. Always 2 for visualization."
    },
    "random_state": {
      "type": "int",
      "default": 0,
      "guidance": "Random seed for reproducibility."
    }
  },
  "outputs": {
    "embedding": "adata.obsm['X_umap']"
  },
  "validation": {
    "no_nan": true,
    "check": "No NaN values in UMAP embedding."
  },
  "notes": "UMAP is for VISUALIZATION ONLY. Never cluster on UMAP coordinates. Never interpret distances between clusters as biologically meaningful. The topology is informative but metric distances are not preserved.",
  "provenance_captures": ["min_dist", "spread", "n_components", "random_state"]
}