Replicability of bulk RNA-Seq differential expression and enrichment analysis results for small cohort sizes

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• RNA sequencing (RNA-Seq) data is high-dimensional and heterogeneous, complicating downstream analyses like differential expression and enrichment.
• Small biological replicate numbers in RNA-Seq experiments, due to practical and financial limits, raise concerns about research replicability, especially in preclinical cancer studies.

Purpose of the Study:

• To investigate the impact of population heterogeneity and small cohort sizes on the replicability of RNA sequencing research.
• To develop a method for researchers to estimate the performance of their RNA-Seq data, particularly when constrained by small sample sizes.

Main Methods:

• Analysis of 18,000 subsampled RNA-Seq experiments using real gene expression data from 18 distinct datasets.
• Development and validation of a bootstrapping procedure to correlate with observed replicability and precision metrics.

Main Results:

• Underpowered RNA-Seq experiments (few replicates) show poor replicability in differential expression and enrichment analyses.
• Low replicability does not always mean low precision; some datasets achieve high median precision even with limited replicates (e.g., >5 replicates).
• The proposed bootstrapping procedure shows strong correlation with actual replicability and precision.

Conclusions:

• Small cohort sizes in RNA-Seq studies significantly hinder the replicability of findings.
• A practical bootstrapping method is provided to help researchers anticipate the performance of their underpowered RNA-Seq datasets.
• Recommendations are offered to mitigate issues associated with underpowered RNA-Seq studies.