Differential gene expression analysis for multi-subject single-cell RNA-sequencing studies with aggregateBioVar

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) offers high resolution but presents challenges in modeling biological variation.
• Bulk RNA sequencing is cost-effective for larger studies, but scRNA-seq is becoming more accessible.
• Accurate statistical modeling of biological variability in scRNA-seq data is crucial but complex for many users.

Purpose of the Study:

• To propose a statistical model for scRNA-seq gene counts.
• To develop a method for estimating model parameters.
• To demonstrate the impact of unmodeled biological variation on false discovery rates (FDRs) in differential expression analysis.

Main Methods:

• Developed a statistical model for scRNA-seq gene counts.
• Implemented a parameter estimation method.
• Evaluated differential expression testing methods using simulation studies and real scRNA-seq datasets (human and animal models).
• Utilized pseudobulk counts as a strategy for improved FDR control.

Main Results:

• A naive differential expression analysis approach inflates FDR when gene expression varies between subjects.
• Differential expression testing on human and animal scRNA-seq data confirmed that naive methods lead to false discoveries.
• An approach using pseudobulk counts demonstrated superior FDR control compared to naive methods.

Conclusions:

• Properly modeling biological variation is essential for accurate scRNA-seq differential expression analysis.
• Ignoring biological variation can lead to a high number of false positives.
• The pseudobulk approach offers a practical solution for controlling FDR in scRNA-seq studies.