Sample size calculations for the differential expression analysis of RNA-seq data using a negative binomial...

Area of Science:

• Genomics and Bioinformatics
• Statistical Genetics
• Biomarker Discovery

Background:

• High-throughput RNA sequencing (RNA-seq) is crucial for disease-related biomarker studies.
• Negative binomial distribution is commonly used for RNA-seq read counts due to over-dispersion.
• Accurate sample size estimation is vital for robust RNA-seq experimental design.

Purpose of the Study:

• To propose two novel, explicit sample size calculation methods for RNA-seq data.
• To utilize a negative binomial regression model incorporating dispersion parameters and size factors.
• To provide a computationally efficient method for experimental design.

Main Methods:

• Developed sample size formulas based on a negative binomial regression model.
• Incorporated common dispersion parameter and size factor with a natural logarithm link function.
• Employed a two-sided Wald test statistic for gene significance testing at FDR 0.05.

Main Results:

• Proposed two new sample size calculation methods for RNA-seq studies.
• Evaluated performance via simulation studies, comparing with existing methods.
• Identified the M1 method as computationally efficient and suitable for quick estimation.

Conclusions:

• The proposed methods provide explicit sample size calculations for RNA-seq experiments.
• The M1 method is recommended for its computational efficiency in experimental design.
• Demonstrated sample size estimation using real-world breast cancer RNA-seq data.