A two-parameter generalized Poisson model to improve the analysis of RNA-seq data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• RNA-sequencing (RNA-seq) is crucial for transcriptome characterization and quantification.
• RNA-seq data analysis faces challenges in managing sequencing bias and normalization.
• Understanding read count distribution is fundamental for accurate RNA-seq analysis.

Purpose of the Study:

• To address challenges in RNA-seq data analysis by modeling position-level read counts.
• To propose and validate a novel statistical model for RNA-seq data.
• To enhance the accuracy of gene expression estimation, normalization, and differential analysis.

Main Methods:

• Development of a two-parameter generalized Poisson (GP) model for position-level read counts.
• Comparison of the GP model's fit against the traditional Poisson model using RNA-seq data.
• Application of the GP model to improve gene expression estimation, normalization, and differential expression/splicing analysis.

Main Results:

• The generalized Poisson (GP) model demonstrates a significantly better fit to position-level RNA-seq read count data compared to the standard Poisson model.
• The GP model facilitates more accurate estimation of gene and exon expression levels.
• Improved normalization strategies and enhanced identification of differentially expressed genes and spliced exons were achieved using the GP model.

Conclusions:

• The generalized Poisson (GP) model offers a superior statistical framework for analyzing RNA-seq data.
• This model improves key aspects of RNA-seq analysis, including expression quantification, normalization, and differential analyses.
• The GP model provides a robust foundation for advancing RNA-seq data interpretation and discovery.