cnmfsns 1.5.0

cNMF Solution Network Space

---

cNMF-SNS: powerful factorization-based multi-omics integration toolkit

Authors: Ted Verhey, Heewon Seo, Sorana Morrissy

cNMF-SNS (consensus Non-negative Matrix Factorization Solution Network Space) is a Python package enabling mosaic integration of bulk, single-cell, and spatial expression data between and within datasets. Datasets can have partially overlapping features (eg. genes) as well as non-overlapping features. cNMF provides a robust, unsupervised deconvolution of each dataset into gene expression programs (GEPs). Network-based integration of GEPs enables flexible integration of many datasets across assays (eg. Protein, RNA-Seq, scRNA-Seq, spatial expression) and patient cohorts.

Communities with GEPs from multiple datasets can be annotated with dataset-specific annotations to facilitate interpretation.

⚡Main Features

Here are just a few of the things that cNMF-SNS does well:

• Identifies interpretable, additive non-negative gene expression programs
• Mosaic integration does not require subsetting features/genes to a shared or overdispersed subset
• Ideal for incremental integration (adding datasets one at a time) since deconvolution is performed independently on each dataset generating invariant GEPs
• Integration does not require comparable data sparsity or sequencing depth (single-cell, microarray, bulk)
• Two interfaces: command-line interface for rapid data exploration and python interface for extensibility and flexibility

🔧 Install

☁️ Public Release

Install the package with conda:

conda install -c conda-forge cnmfsns

✨ Latest version from GitHub

Before installing cNMF-SNS using pip, it is recommended to first set up a separate conda environment and have conda manage as many dependencies as possible.

conda create --name cnmfsns -c conda-forge python=3.10 anndata pandas numpy scipy matplotlib upsetplot httplib2 tomli tomli-w click pygraphviz python-igraph semantic_version pyyaml scikit-learn fastcluster scanpy pyyaml
conda activate cnmfsns
pip install git+https://github.com/MorrissyLab/cNMF-SNS.git

📖 Documentation

🗐 Data requirements

cNMF-SNS can factorize a wide variety of datasets, but will work optimally in these conditions:

• Use untransformed (raw) data where possible. For (sc)RNA-Seq data, use feature counts, not TPM or other log-transformed values.

📓 Python interface

To get started, sample proteomics datasets and a Jupyter notebook tutorial is available here.

Detailed API reference can be found on ReadTheDocs.

⌨️ Command line interface

See the command line interface documentation.

💭 Getting Help

For errors arising during use of cNMF-SNS, create and browse issues in the GitHub "issues" tab.