carmonalab / GeneNMF

Methods to discover gene programs on single-cell data
71 stars 3 forks source link

GeneNMF: unsupervised discovery of gene programs in single-cell data

Non-negative matrix factorization is a method for the analysis of high dimensional data that allows extracting sparse and meaningful features from a set of non-negative data vectors. It is well suited for decomposing scRNA-seq data, effectively reducing large complex matrices ($10^4$ of genes times $10^5$ of cells) into a few interpretable gene programs. It has been especially used to extract recurrent gene programs in cancer cells (see e.g. Barkely et al. (2022) and Gavish et al. (2023)), which are otherwise difficult to integrate and analyse jointly.

GeneNMF is a package that implements methods for matrix factorization and gene program discovery for single-cell omics data. It can be applied directly on Seurat objects to reduce the dimensionality of the data and to detect robust gene programs across multiple samples. For fast NMF calculation, GeneNMF relies on RcppML (see DeBruine et al. 2024).

Installation

Install release version from CRAN:

install.packages("GeneNMF")

Or for the latest version, install from GitHub:

library(remotes)
remotes::install_github("carmonalab/GeneNMF")

Test your installation

library(GeneNMF)
data(sampleObj)
sampleObj <- runNMF(sampleObj, k=5)

Meta programs discovery using default parameters

Perform NMF over a list of Seurat objects and for multiple values of k (number of NMF factors) to extract gene programs

sampleObj.list <- Seurat::SplitObject(sampleObj, split.by = "donor")
geneNMF.programs <- multiNMF(sampleObj.list, k=4:9)

Cluster gene programs from multiple samples and k's into meta-programs (MPs), i.e. consensus programs that are robustly identified across NMF runs. Compute MP metrics and most influencial MP genes.

geneNMF.metaprograms <- getMetaPrograms(geneNMF.programs, nMP=5)

GeneNMF demos

Find demos of the functionalities of GeneNMF and more explanations in the following tutorials:

For the source code, see the GeneNMF.demo repository.

News: version 0.6 is here

We made some improvements to the algorithm to allow more robust identification of metaprograms, and easier tuning of parameters. Here are the main changes:

Citation

If you used GeneNMF in your work, please cite:

Wounding triggers invasive progression in human basal cell carcinoma. Laura Yerly, Massimo Andreatta, Josep Garnica, Jeremy Di Domizio, Michel Gilliet, Santiago J Carmona, Francois Kuonen. bioRxiv 2024 10.1101/2024.05.31.596823