log_normalize.json

{
  "tool_id": "log_normalize",
  "name": "Log Normalization (CP10K + log1p)",
  "category": "normalization",
  "framework": "scanpy",
  "function": "sc.pp.normalize_total + sc.pp.log1p",
  "valid_after": [
    "qc",
    "qc"
  ],
  "valid_before": [
    "feature_selection"
  ],
  "paradigms": [
    "all"
  ],
  "parameters": {
    "target_sum": {
      "type": "float",
      "default": 10000,
      "guidance": "Normalize each cell to this total count before log-transforming. 1e4 (CP10K) is the standard. Some workflows use 1e6 (CPM) but CP10K is recommended for scRNA-seq."
    },
    "exclude_highly_expressed": {
      "type": "bool",
      "default": false,
      "guidance": "If true, exclude very highly expressed genes from the normalization factor. Can help when a few genes dominate total counts."
    }
  },
  "outputs": {
    "normalized_adata": "AnnData with log-normalized values in adata.X",
    "raw_counts": "Original raw counts preserved in adata.raw"
  },
  "validation": {
    "values_are_log_scale": true,
    "max_value_lt": 15,
    "raw_exists": true,
    "check": "adata.X values should be log-scale (max typically < 12). adata.raw must exist for downstream DE testing."
  },
  "provenance_captures": [
    "target_sum",
    "exclude_highly_expressed"
  ]
}