Related Experiment Video
Updated: Sep 6, 2025
10:10
Three Differential Expression Analysis Methods for RNA Sequencing: limma, EdgeR, DESeq2
Published on: September 18, 2021
38.4K
Recommendations of scRNA-seq Differential Gene Expression Analysis Based on Comprehensive Benchmarking
Jake Gagnon1, Lira Pi2, Matthew Ryals2
1Analytics and Data Sciences, Biogen, Inc., 225 Binney St., Cambridge, MA 02142, USA.
Life (Basel, Switzerland)
|June 24, 2022
Summary
We developed a novel simulator for single-cell RNA sequencing (scRNA-seq) data to benchmark differential gene expression (DGE) analysis tools. Negative binomial mixed models, like NEBULA-HL, proved superior for identifying cell-type-specific transcriptomic responses.
Area of Science:
- Genomics
- Bioinformatics
- Computational Biology
Background:
- Single-cell RNA sequencing (scRNA-seq) generates complex data with multiple sources of variation.
- Selecting appropriate tools and parameters for differential gene expression (DGE) analysis in scRNA-seq is challenging for researchers.
- Accurate DGE analysis is crucial for understanding biological responses to disease or treatment.
Purpose of the Study:
- To develop a novel simulator for scRNA-seq data that accurately reflects real-world biological and technical variations.
- To benchmark and compare the performance of 12 different DGE analysis methods using simulated multi-subject, multi-condition data.
- To provide recommendations for optimal cell and gene filtering strategies to enhance DGE analysis outcomes.
Main Methods:
- Development of a novel simulator for scRNA-seq data, incorporating cell-to-cell, subject, cell type, gene expression, library size, group, and covariate effects.
- Benchmarking of 12 DGE analysis methods, including cell-level and pseudo-bulk approaches, on simulated 10x Genomics data.
- Application of the developed pipeline and methods to two real scRNA-seq datasets.
Main Results:
- Methods based on the negative binomial mixed model, specifically glmmTMB and NEBULA-HL, demonstrated superior performance in DGE analysis.
- The NEBULA-HL method, integrated into a statistical analysis pipeline, effectively identified cell-type-specific transcriptomic responses.
- The study confirmed the outperformance of the proposed DGE pipeline on real datasets, yielding consistent differentially expressed gene (DEG) findings and pseudo-time trajectory results.
Conclusions:
- The developed scRNA-seq simulator is a valuable tool for evaluating DGE analysis methods.
- Negative binomial mixed models, particularly NEBULA-HL, are recommended for robust DGE analysis in complex scRNA-seq experiments.
- The findings facilitate better understanding of transcriptomic changes in disease/treatment and aid in identifying new drug targets.
