Confronting false discoveries in single-cell differential expression.

Area of Science:

• Single-cell transcriptomics
• Computational biology
• Bioinformatics

Background:

• Differential expression analysis is key to understanding cell-type specific responses in single-cell RNA sequencing (scRNA-seq).
• Numerous statistical methods exist for identifying differentially expressed genes (DEGs), but their performance and underlying principles are not well understood.
• Variation between biological replicates is an inherent challenge in scRNA-seq data analysis.

Purpose of the Study:

• To evaluate the performance of various differential expression analysis methods in single-cell transcriptomics.
• To identify the key principles that differentiate method performance.
• To highlight the impact of biological variability on DEG detection.

Main Methods:

• Comparative analysis of statistical methods for DEG identification in scRNA-seq data.
• Assessment of method sensitivity to variation between biological replicates.
• Demonstration of true and false DEG discoveries using a mouse spinal cord injury model.

Main Results:

• Many widely used DEG analysis methods are biased and produce false discoveries when biological variation is not properly accounted for.
• Methods ignoring inter-replicate variability can identify hundreds of DEGs even without true biological differences.
• The study exemplifies these findings using data from an injured mouse spinal cord.

Conclusions:

• Accurate differential expression analysis in single-cell transcriptomics necessitates methods that rigorously account for biological variability.
• Failure to address inter-replicate variation leads to unreliable DEG identification and inflated false discovery rates.
• The findings underscore the importance of method selection for robust biological interpretation of scRNA-seq data.