Skip to content

How to configure metadata curation

This guide explains how to apply curation operations to cell metadata, feature metadata, or biosample metadata during a single-cell HDF5 transformation.

Where these operations apply

Curation operations are available in the cell_metadata, feature_metadata, and per-entity settings within biosample_metadata. They are applied in the order listed below.

Order of operations

1. Drop columns

Remove columns before any other processing:

"columns_to_drop": ["taxon", "organism_id"]

2. Rename columns

Map source column names to new names:

"columns_renaming_map": {
  "sample": "batch",
  "pctmt": "percentMito"
}

3. Replace specific values

Replace known values within a column:

"columns_to_curate_values": {
  "sample": {
    "LGVXCTRL1": "lung_healthy_1"
  }
}

4. Fill missing values

Provide a default value for missing entries:

"columns_to_fill_missing_values": {
  "batch": "unknown"
}

5. Set a constant value for all rows

Set all rows in a column to the same value. This can add new columns or overwrite existing ones:

"set_column_value": {
  "sample_id": "lung_1"
}

Attribute name standardisation

After all explicit column operations, the transformation applies automatic attribute name standardisation: column names that match known ODM canonical names are mapped to those names; non-standard names are converted to camelCase. This step is automatic and does not need to be configured.

For the full list of recognised column names and their ODM equivalents, see Attribute Mapping Reference.

Exempt a column from standardisation

To prevent a specific column from being automatically renamed (for example, a Leiden cluster column with a decimal suffix), list it in columns_to_preserve_name:

"columns_to_preserve_name": ["cluster_leiden_0.5"]