README.md

In CellDEEP: Cell DiffErential Expression by Pooling ('CellDEEP')

CellDEEP

Introduction to CellDEEP

Single-cell RNA sequencing (scRNA-seq) allows us to explore gene expression at an unprecedented resolution, but it faces significant challenges like high dropout rates and data sparsity. CellDEEP was developed to bridge the gap between robust but coarse pseudobulk methods and sensitive but potentially biased single-cell methods.

CellDEEP overview

Check our paper for details: https://www.biorxiv.org/content/10.64898/2026.03.09.710522v1

Installation

You can install the development version of CellDEEP via github:

devtools::install_github("sii-scRNA-Seq/CellDEEP")

Example

Before using, don’t forget library it:

library(CellDEEP)

Vignette:

• Tutorial vignette

To quickly run CellDEEP, pass your metadata column names directly into FindMarker.CellDEEP:

data("sim")

# Pool defaults to TRUE
de.test <- FindMarker.CellDEEP(sim, 
                          group_id = "Status", 
                          sample_id = "DonorID", 
                          cluster_id = "cluster_id",
                          Pool = TRUE,
                          test.use = "wilcox", 
                          n_cells = 3, 
                          min_cells_per_subgroup = 1,
                          cell_selection = "random", 
                          readcounts = "sum", 
                          logfc.threshold = 0.25, 
                          ident.1 = "Case", 
                          ident.2 = "Control")

This section introduce what is updated.

For version 1.0.1: 1. FindMarker.CellDEEP Pool default should be TRUE. 2. Not clear what is cell_cutoff, replaced with new parameter 3. Change “pool_way” to cell_selection 4. Vignette easy to access/read 5. Change toy data to simulated data, will generate DE result now.

For publish version(1.0.0): Delete code used for experiment, keep only CellDEEP function code. Delete all the comments, clean the code. Rename pooling function as CellDEEP.Kmean and CellDEEP.Random.

CellDEEP documentation built on March 29, 2026, 5:08 p.m.