# Mowgli: Multi Omics Wasserstein inteGrative anaLysIs

[![Tests](https://github.com/gjhuizing/Mowgli/actions/workflows/main.yml/badge.svg)](https://github.com/gjhuizing/Mowgli/actions/workflows/main.yml)

[![codecov](https://codecov.io/gh/cantinilab/Mowgli/branch/main/graph/badge.svg?token=UBUJF7098Q)](https://codecov.io/gh/cantinilab/Mowgli)

[![Documentation Status](https://readthedocs.org/projects/mowgli/badge/?version=latest)](https://mowgli.readthedocs.io/en/latest/?badge=latest)

[![PyPI version](https://img.shields.io/pypi/v/mowgli?color=blue)](https://img.shields.io/pypi/v/mowgli?color=blue)

[![Code style: black](https://img.shields.io/badge/code%20style-black-000000.svg)](https://github.com/psf/black)

[![DOI](https://zenodo.org/badge/391909874.svg)](https://zenodo.org/badge/latestdoi/391909874)

Mowgli is a novel method for the integration of paired multi-omics data with any type and number of omics, combining integrative Nonnegative Matrix Factorization and Optimal Transport. [Read the paper!](https://www.nature.com/articles/s41467-023-43019-2)

![figure](figure.png)

## Install the package

Mowgli is implemented as a Python package seamlessly integrated within the scverse ecosystem, in particular Muon and Scanpy.

### via PyPI (recommended)

On all operating systems, the easiest way to install Mowgli is via PyPI. Installation should typically take a minute and is continuously tested with Python 3.10 on an Ubuntu virtual machine.

```bash

pip install mowgli

```

### via GitHub (development version)

```bash

git clone git@github.com:cantinilab/Mowgli.git

pip install ./Mowgli/

```

### Test your installation (optional)

```bash

pytest .

```

## Getting started

Mowgli takes as an input a Muon object and populates its `obsm` and `uns` fields with the embeddings and dictionaries, respectively. Visit [mowgli.rtfd.io](https://mowgli.rtfd.io/) for more documentation and tutorials.

You may download a preprocessed 10X Multiome demo dataset [here](https://figshare.com/s/4c8e72cbb188d8e1cce8).

A GPU is not required for small datasets, but is strongly recommended above 1,000 cells. On CPU, the [cell lines demo](https://mowgli.readthedocs.io/en/latest/vignettes/Liu%20cell%20lines.html) (206 cells) should run in under 5 minutes and the [PBMC demo](https://mowgli.readthedocs.io/en/latest/vignettes/PBMC.html) (500 cells) should run in under 10 minutes (tested on a Ubuntu 20.04 machine with an 11th gen i7 processor).

```python

import mowgli

import mudata as md

import scanpy as sc

# Load data into a Muon object.

mdata = md.read_h5mu("my_data.h5mu")

# Initialize and train the model.

model = mowgli.models.MowgliModel(latent_dim=15)

model.train(mdata)

# Visualize the embedding with UMAP.

sc.pp.neighbors(mdata, use_rep="W_OT")

sc.tl.umap(mdata)

sc.pl.umap(mdata)

```

## Publication

```bibtex

@article{huizing2023paired,

  title={Paired single-cell multi-omics data integration with Mowgli},

  author={Huizing, Geert-Jan and Deutschmann, Ina Maria and Peyr{\'e}, Gabriel and Cantini, Laura},

  journal={Nature Communications},

  volume={14},

  number={1},

  pages={7711},

  year={2023},

  publisher={Nature Publishing Group UK London}

}

```

If you're looking for the repository with code to reproduce the experiments in our preprint, [here is is!](https://github.com/cantinilab/mowgli_reproducibility)