---
id: mudata
version: "0.3.10"
license: BSD 3-Clause License Copyright (c) 2020, Danila Bredikhin Copyright (c) 2025, scverse® All rights reserved. Redistribution and use in source and binary forms, with or without modification, are… (full text in the JSON record)
license_treatment: permissive
maintenance: active
---

# mudata — Multimodal data

License: permissive · Maintenance: active · Downloads: 194.2K/mo

## What it is and what it does

MuData extends the AnnData single-cell data structure to handle multimodal omics experiments where multiple assays (RNA, ATAC, protein, etc.) are measured on the same cells. It stores each modality as a separate AnnData object within a unified container, allowing shared cell and feature metadata while preserving modality-specific annotations. The package provides HDF5-based persistence through .h5mu files, which mirror the .h5ad format but support hierarchical storage of multiple modalities.

Typically used in single-cell genomics workflows where data from technologies like 10X Genomics multiome (RNA + ATAC) or other multi-assay platforms need to be loaded, processed, and stored together. The package is part of the scverse ecosystem and integrates with muon for convenient readers of common multimodal file formats. It depends on anndata, h5py, numpy, pandas, and scipy for core functionality.

Use it for:

- Store and retrieve 10X Genomics multiome data (RNA + ATAC) with shared cell metadata and modality-specific feature annotations.
- Combine multiple single-cell assays measured on the same cells into one analysis-ready container with linked observations.
- Save multimodal analysis results to disk in a structured HDF5 format and selectively read individual modalities back.
- Organize metadata for multimodal experiments where cell-level annotations apply across all assays but feature annotations differ per modality.
- Build reproducible multimodal analysis pipelines that preserve data structure and metadata through multiple processing steps.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

MuData is a container for multimodal omics data that organizes multiple AnnData objects (one per modality) with shared and modality-specific observations and features, and provides HDF5-based I/O via .h5mu files.

Yes. MuData is actively maintained, has no known vulnerabilities, and fills a clear role in the scverse ecosystem for multimodal omics data. The low install friction, permissive license, and integration with established tools like anndata and muon make it a straightforward choice if you work with multimodal single-cell data. Install it if you need to organize or persist data from multiple assays measured on the same cells.

## Install

pip install mudata
uv add mudata
poetry add mudata

## Installing mudata

Before you install: Low install friction with a pure-wheel distribution. Active maintenance with a release 38 days ago and recent commits. Seven runtime dependencies are all established scientific Python packages (anndata, h5py, numpy, pandas, scipy, scverse-misc, session-info2).

License in practice: BSD 3-Clause permissive license allows commercial and private use with attribution and liability disclaimer. No restrictions on modification or redistribution.

Quickstart:

pip install mudata

from mudata import MuData

mdata = MuData({'rna': adata_rna, 'atac': adata_atac})
mdata.write('multimodal.h5mu')
mdata = MuData.read('multimodal.h5mu')

Requires Python 3.10 or later. Assumes anndata objects (adata_rna, adata_atac) are already available.

Verify before relying:

- Whether the package handles edge cases like mismatched observation counts across modalities gracefully.
- Performance characteristics when working with very large multimodal datasets (millions of cells, thousands of features per modality).
- Interoperability with tools outside the scverse ecosystem for reading/writing .h5mu files.

## Package facts

- License: BSD 3-Clause License Copyright (c) 2020, Danila Bredikhin Copyright (c) 2025, scverse® All rights reserved. Redistribution and use in source and binary forms, with or without modification, are… (full text in the JSON record) (permissive)
- Python support: supports_current
- Install friction: low
- Maintenance: active
- Downloads: 194.2K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

multimodal omics data structure, multi-assay single-cell data, anndata container for multiple modalities, h5mu file format reader writer, scverse multimodal analysis, single-cell-genomics, multimodal-data, scverse

[View on SkillFed](https://skillfed.io/packages/mudata) · [View on PyPI](https://pypi.org/project/mudata/)