A new shrinkage estimator for dispersion improves differential expression detection in RNA-seq data

Area of Science:

• Genomics
• Bioinformatics
• Statistical Genetics

Background:

• RNA-sequencing (RNA-seq) generates extensive gene expression count data.
• Differential expression (DE) analysis is crucial for functional genomics.
• Current DE methods often use the negative binomial model but vary in dispersion estimation.

Purpose of the Study:

• To evaluate the adequacy of dispersion estimation in existing RNA-seq DE methods.
• To develop an improved method for estimating dispersion parameters in RNA-seq data.
• To enhance the accuracy of differential gene expression detection.

Main Methods:

• Evaluation of several leading DE detection methods using large public RNA-seq datasets.
• Development of a novel empirical Bayes shrinkage estimation for dispersion parameters.
• Comparison of the new method's performance against existing approaches.

Main Results:

• Existing dispersion estimation methods inadequately capture biological variance heterogeneity in RNA-seq data.
• The proposed empirical Bayes shrinkage method provides a more accurate dispersion estimate.
• Improved DE detection rates were demonstrated using the new estimation approach.

Conclusions:

• Accurate modeling of biological variance is critical for robust RNA-seq DE analysis.
• The novel empirical Bayes shrinkage estimator offers a significant advancement in DE detection.
• This method enhances the reliability of functional genomics studies relying on RNA-seq data.