Skip to content

How to process a 10x Genomics H5 file

This guide explains how to ingest a 10x Genomics H5 file through the ODM single-cell transformation pipeline.

Required configuration change

Set file_type to "h5" instead of "h5ad". Use the same H5AD key names (obs, var) in metadata_keys: the transformation converts the 10x H5 format to H5AD internally before applying unified processing.

{
  "file_type": "h5",
  "cell_metadata": {
    "metadata_keys": {
      "obs": "metadata"
    }
  },
  "feature_metadata": {
    "metadata_keys": {
      "var": "metadata"
    }
  },
  "cell_expression": {
    "data_class": "Single-cell transcriptomics"
  }
}

Volume sizing

When setting volume_size for a job using an H5 input file, allocate at least 4× the original attachment size (for example, a 5 GB file requires volume_size ≥ 20 GB). H5 inputs require additional scratch space because the transformation converts them to H5AD during processing.

Legacy 10x H5 support

Legacy 10x Genomics H5 files (v<3) are supported only when the file contains a single genome. If the file includes multiple genomes, pre-process it to extract the genome of interest before running the transformation.

Prerequisites and workflow

For the full job submission workflow, see How to run a transformation. For authentication prerequisites, see Authentication and tokens.