muon 0.1.9

muon is a multimodal omics Python framework.

Documentation | Tutorials | Publication

Data structure

muon is designed around MuData (multimodal data) objects — in the same vein as scanpy and AnnData are designed to work primarily with scRNA-seq data in Python. Individual modalities in MuData are naturally represented with AnnData objects.

MuData class and .h5mu files I/O operations are part of the standalone mudata library.

Input

MuData class is implemented in the mudata library and is exposed in muon:

from muon import MuData

mdata = MuData({'rna': adata_rna, 'atac': adata_atac})

If multimodal data from 10X Genomics is to be read, muon provides a reader that returns a MuData object with AnnData objects inside, each corresponding to its own modality:

import muon as mu

mu.read_10x_h5("filtered_feature_bc_matrix.h5")
# MuData object with n_obs × n_vars = 10000 × 80000 
# 2 modalities
#   rna:	10000 x 30000
#     var:	'gene_ids', 'feature_types', 'genome', 'interval'
#   atac:	10000 x 50000
#     var:	'gene_ids', 'feature_types', 'genome', 'interval'
#     uns:	'atac', 'files'

I/O with .h5mu files

Basic .h5mu files I/O functionality is implemented in mudata and is exposed in muon. A MuData object represents modalities as collections of AnnData objects, and these collections can be saved on disk and retrieved using HDF5-based .h5mu files, which design is based on .h5ad file structure.

mdata.write("pbmc_10k.h5mu")
mdata = mu.read("pbmc_10k.h5mu")

It allows to effectively use the hierarchical nature of HDF5 files and to read/write AnnData object directly from/to .h5mu files:

adata = mu.read("pbmc_10k.h5mu/rna")
mu.write("pbmc_10k.h5mu/rna", adata)

Multimodal omics analysis

muon incorporates a set of methods for multimodal omics analysis. These methods address the challenge of taking multimodal data as their input. For instance, while for a unimodal analysis one would use principal components analysis, muon comes with a method to run multi-omics factor analysis:

# Unimodal
import scanpy as sc
sc.tl.pca(adata)

# Multimodal
import muon as mu
mu.tl.mofa(mdata)

Individual assays

Individual assays are stored as AnnData object, which enables the use of all the default scanpy functionality per assay:

import scanpy as sc

sc.tl.umap(mdata.mod["rna"])

Typically, a modality inside a container can be referred to with a variable to make the code more concise:

rna = mdata.mod["rna"]
sc.pl.umap(rna)

Modules in muon

muon comes with a set of modules that can be used hand in hand with scanpy's API. These modules are named after respective sequencing protocols and comprise special functions that might come in handy. It is also handy to import them as two letter abbreviations:

# ATAC module:
from muon import atac as ac

# Protein (epitope) module:
from muon import prot as pt

Some implementation details are noted in DESIGN.md.

Contributions in the form of issues, pull requests or discussions are welcome.

Citation

If you use muon in your work, please cite the muon publication as follows:

MUON: multimodal omics analysis framework

Danila Bredikhin, Ilia Kats, Oliver Stegle

Genome Biology 2022 Feb 01. doi: 10.1186/s13059-021-02577-8.

You can cite the scverse publication as follows:

The scverse project provides a computational ecosystem for single-cell omics data analysis

Isaac Virshup, Danila Bredikhin, Lukas Heumos, Giovanni Palla, Gregor Sturm, Adam Gayoso, Ilia Kats, Mikaela Koutrouli, Scverse Community, Bonnie Berger, Dana Pe’er, Aviv Regev, Sarah A. Teichmann, Francesca Finotello, F. Alexander Wolf, Nir Yosef, Oliver Stegle & Fabian J. Theis

Nat Biotechnol. 2023 Apr 10. doi: 10.1038/s41587-023-01733-8.

muon is part of the scverse® project (website, governance) and is fiscally sponsored by NumFOCUS. If you like scverse® and want to support our mission, please consider making a tax-deductible donation to help the project pay for developer time, professional services, travel, workshops, and a variety of other needs.