Variability in estimated gene expression among commonly used RNA-seq pipelines.

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• RNA-sequencing (RNA-seq) is crucial for identifying disease biomarkers and therapeutic targets.
• Accurate mRNA abundance estimation is fundamental for downstream RNA-seq analyses.
• Current RNA-seq processing pipelines are assumed to provide reliable expression level estimates.

Purpose of the Study:

• To evaluate the consistency of gene expression abundance estimates across different RNA-seq processing pipelines.
• To identify protein-coding genes with significant expression discrepancies among pipelines.
• To assess the impact of pipeline variability on disease-associated gene identification.

Main Methods:

• Analysis of RNA-seq data from 6,690 human tumor and normal tissue samples.
• Application of five distinct, best-in-class RNA-seq processing pipelines to identical datasets.
• Comparison of mRNA abundance estimates and expression fold changes across pipelines for >19,000 protein-coding genes.

Main Results:

• Approximately 88% of protein-coding genes exhibited similar expression profiles across all tested pipelines.
• Over 12% of protein-coding genes showed more than four-fold differences in abundance estimates between pipelines.
• Discordant gene expression profiles were observed for many widely studied disease-associated genes.
• Variability in abundance estimates was influenced by diverse patterns of inter-pipeline discordance.

Conclusions:

• Significant pipeline-dependent variability exists in RNA-seq abundance estimates for a notable subset of genes.
• This variability can affect the reliability of disease biomarker and therapeutic target discovery.
• A community-wide effort is required to establish gold standards for estimating mRNA abundance for discordant genes.
• The identified list of discordantly evaluated genes serves as a resource for more robust marker discovery.