GmGM 0.4.0

An implementation of the Gaussian multi-Graphical Model

GmGM-python

A python package for the GmGM algorithm. Read the pre-print here.

This is a very early version so the API is subject to change.

Installation

We recommend installing this package in a conda environment, by first running:

conda create -n {YOUR ENVIRONMENT NAME} "python>=3.9,<3.12"
conda activate {YOUR ENVIROMENT NAME}

Afterwards you can install it via pip.

# Pip
python -m pip install GmGM

Conda install coming soon.

About

This package learns a graphical representation of every "axis" of your data. For example, if you had a paired scRNA+scATAC multi-omics dataset, then your axes would be "genes" (columns of scRNA matrix), "axes" (columns of scATAC matrix), and "cells" (rows of both matrices).

This package works on any dataset that can be expressed as multiple tensors of arbitrary length (so multi-omics, videos, etc...). The only restriction is that no tensor can have the same axis twice (no "genes x genes" matrix); the same axis can appear multiple times, as long as it only appears once per matrix.

Usage

For an example, we recommend looking at the danio_rerio.ipynb notebook.

With AnnData

If you already have your data stored as an AnnData object, GmGM can be used directly. Suppose you had a single-cell RNA sequencing dataset scRNA.

GmGM(
    scRNA,
    to_keep={
        "obs": 10,
        "var": 10,
    }
)

"obs": 10 tells the algorithm to keep 10 edges per cell (the 'obs' axis of AnnData) and "var": 10 tells the algorithm to keep 10 edges per gene (the 'var' axis of AnnData).

This modifies the AnnData object in place, storing the resultant graphs in scRNA.obsp["obs_gmgm_connectivities"] and scRNA.varp["var_gmgm_connectivities"].

With MuData

Mudata support is very similar to AnnData support. Suppose we had a MuData object mudata with scATAC and scRNA data, then:

GmGM(
    mudata,
    to_keep={
        "obs": 10,
        "rna-var": 10,
        "atac-var": 10
    }
)
# Cell graph
scRNA.obsp["obs_gmgm_connectivities"]
# Gene graph
scRNA.varp["rna-var_gmgm_connectivities"]
# Peak graph
scRNA.varp["atac-var_gmgm_connectivities"]

In general, accessing features can be done by appending the name of the modality onto "var", i.e. "metabolomics-var" if the MuData has a metabolomics modality.

Note that we (will, not currently) support MuData with the axis=1 and axis=-1 parameters as well.

General Usage (i.e. Without AnnData/MuData)

The first step is to express your dataset as a Dataset object. Suppose you had a cells x genes scRNA matrix and cells x peaks scATAC matrix, then you could create a Dataset object like:

from GmGM import Dataset
dataset: Dataset = Dataset(
    dataset={
        "scRNA": scRNA,
        "scATAC": scATAC
    },
    structure={
        "scRNA": ("cell", "gene"),
        "scATAC": ("cell", "peak")
    }
)

Running GmGM is then as simple as:

GmGM(
    dataset,
    to_keep={
        "cell": 10,
        "gene": 10,
        "peak": 10
    }
)

to_keep tells the algorithm how many edges to keep per cell/gene/peak.

The final results are stored in dataset.precision_matrices["cell"], dataset.precision_matrices["gene"], and dataset.precision_matrices["peak"], respectively.

Roadmap

• Add direct support for AnnData objects
• Add direct support for MuData objects (so that converson to Dataset is not needed)
• Stabilize API
• Add comprehensive docs
• Have generate_data directly generate Dataset objects
• Add conda distribution
• Add example notebook
• Make sure regularizers still work
• Make sure priors still work
• Make sure covariance thresholding trick still works
• Add unit tests
• Allow forcing subset of axes to have given precision matrices
• Add random generation of count matrix data