Modeling expression ranks for noise-tolerant differential expression analysis of scRNA-Seq data


ROSeq - A rank based approach to modeling gene expression with filtered and normalized read count matrix. ROSeq takes filtered and normalized read matrix and cell-annotation/condition as input and determines the differentially expressed genes between the contrasting groups of single cells. One of the input parameters is the number of cores to be used.


The developer’s version of the R package can be installed with the following R commands:

The github’s version of the R package can be installed with the following R commands:

Vignette tutorial

This vignette uses the Tung dataset, which is already inbuilt in the package, to demonstrate a standard pipeline.


Libraries need to be loaded before running.

Data Preprocessing:

Cells and genes filtering then voom transformation after TMM normalization

Below commands can be used for Cell/gene filtering, TMM normalization and voom transformation. The user is free to use an alternative preprocessing strategy while using different filtering/normalization methods.

ROSeq analysis.

Input: gene expression matrix with genes in rows and cells in columns. Condition/group annotation of cells also need to be supplied. User can set numCores based the hardware specifications in her computer.

Showing results are in the form of pVals, pAdj and log2FC

p_Vals : p_value (unadjusted)
p_Adj : Adjusted p-value, based on FDR method

Publication (preprint available at bioRxiv):

Gupta, K., Lalit, M., Biswas, A., Maulik, U., Bandyopadhyay, S., Ahuja, G., Ghosh, A., and Sengupta, D., 2020. ROSeq: Noise-tolerant differential expression analysis of scRNA-Seq data. DOI: