Valid Post-clustering Differential Analysis for Single-Cell RNA-Seq.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Single-cell computational pipelines commonly use clustering to organize cells and differential expression analysis to identify marker genes.
• Current methods often perform differential expression analysis on the same dataset used for clustering.
• This approach can lead to artificially low p-values and false positive marker discoveries due to the inherent "forced" separation introduced by clustering.

Purpose of the Study:

• To address the issue of false discoveries in single-cell differential expression analysis.
• To introduce a statistically valid framework for post-clustering differential analysis.
• To improve the reliability of marker gene identification in single-cell studies.

Main Methods:

• Developed a novel computational framework for differential expression analysis after cell clustering.
• Implemented a statistical approach to correct for biases introduced by the clustering step.
• Validated the framework using simulated and real single-cell datasets.

Main Results:

• The proposed framework significantly reduces false positive marker discoveries compared to standard methods.
• The corrected p-values accurately reflect the true differential expression between cell clusters.
• Software implementing the framework is publicly available.

Conclusions:

• Reusing clustered data for differential expression analysis inflates false discovery rates.
• The introduced framework provides a statistically sound method for accurate marker gene identification.
• This advancement enhances the reliability of single-cell data interpretation and downstream biological insights.