COMSE: Feature selection

Mar 13, 2024 with COMSE version 1.0.0

Abstract

COMSE[1] is an unsupervised feature selection framework using community detection to capture informative genes from scRNA-seq data. COMSE identified homogenous cell substates with high resolution, as demonstrated by distinguishing different cell cycle stages. Evaluations based on real and simulated scRNA-seq datasets showed COMSE outperformed methods even with high dropout rates in cell clustering assignment. We also demonstrate that by identifying communities of genes associated with batch effects, COMSE parses signals reflecting biological difference from noise arising due to differences in sequencing protocols, thereby enabling integrated analysis of scRNA-seq datasets generated differently.

Package

• R: igraph, Rdimtools

Contents

1 Introduction to COMSE

COMSE first partitions all genes into different communities in latent space inferred by Principle Component Analysis (PCA) using the Louvain algorithm[2]. Within each community, we apply a denoising procedure to remove noise introduced during sequencing or other procedures. It then selects highly informative genes from each community based on the Laplacian score [3] (Fig. 1). For more information on COMSE, we recommend the user to check the following article:

COMSE: Analysis of Single-Cell RNA-seq Data Using Community Detection Based Feature Selection (https://doi.org/10.1186/s12915-024-01963-5)

Please cite this article if you use COMSE in your research.

2 Requirements

2.1 Package installation

For R platform, the core dependent packages are igraph and Rdimtools.

install.packages("igraph")
install.packages("Rdimtools")

2.2 Input: expression matrix

The input of COMSE is the expression matrix of scRNA-seq/bulk RNA-seq data:

• Each column represents a cell sample and each row represents a gene.
• The row name of the matrix should be the gene-symbol of gene ID.
• Expression units: The preferred expression values are raw values. Since we will normalize the data by default.

3 COMSE Running demo

library(igraph)
library(Rdimtools)
source("Functions_COMSE_Denoise.R")
## Loading data
data = readRDS("demo_data.RDS")
counts = data[[1]]
meta = data[[2]]
result = COMSE(counts)
result = DenoiseRandom(counts) ## Optional (return denoised data)

Reference

[1] Luo Q, Chen Y, Lan X (2024) COMSE: analysis of single-cell RNA-seq data using community detection-based feature selection. BMC Biol 22:167. https://doi.org/10.1186/s12915-024-01963-5

[2] Blondel,V.D., Guillaume,J.-L., Lambiotte,R. and Lefebvre,E. (2008) Fast unfolding of communities in large networks. Journal of Statistical Mechanics: Theory and Experiment, https://doi.org/10.1088/1742-5468/2008/10/P10008

[3] He,X., Cai,D. and Niyogi,P. (2005) Laplacian Score for Feature Selection. In 18th International Conference on Neural Information Processing Systems.pp. 507–514.