Gene expression variability and the analysis of large-scale RNA-seq studies with the MDSeq

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Decreasing costs of next-generation sequencing (NGS) have enabled large-scale RNA-seq data generation.
• RNA-seq data allows for the analysis of both gene expression means and variability.
• Existing methods may face challenges in analyzing large-scale RNA-seq datasets efficiently and robustly.

Purpose of the Study:

• To introduce MDSeq, a novel computational tool for analyzing gene expression means and variability in RNA-seq data.
• To provide a robust and computationally efficient method for analyzing large-scale RNA-seq datasets.
• To develop a comprehensive toolset addressing challenges such as technical excess zeros and outlier identification.

Main Methods:

• MDSeq is based on the coefficient of dispersion.
• It employs a novel reparametrization of the negative binomial distribution for flexible generalized linear models (GLMs) of mean and dispersion.
• The toolset incorporates methods for modeling technical excess zeros, efficient outlier identification, and differential expression analysis.

Main Results:

• MDSeq demonstrated superior performance compared to current methods in analyzing gene expression means and variability using simulated data.
• Application to real RNA-seq studies identified functionally relevant genes and pathways.
• Analysis of GTEx human brain tissue data revealed pathways associated with neurodegenerative disorders, even when gene expression means remained conserved.

Conclusions:

• MDSeq provides a powerful and efficient approach for analyzing gene expression means and variability in large-scale RNA-seq data.
• The method is capable of identifying biologically relevant insights, including disease-associated pathways, by focusing on expression variability.
• MDSeq represents a significant advancement in the analysis of RNA-seq data for biological discovery.