A Multivariate Negative-Binomial Model with Random Effects for Differential Gene-Expression Analysis of Correlated...

Area of Science:

• Bioinformatics
• Statistical Genomics
• Computational Biology

Background:

• Correlated RNA sequencing (RNA-Seq) data from experimental designs like matched-pair or longitudinal studies are common.
• Existing RNA-Seq analysis methods often struggle with specific designs or require balanced data, limiting their applicability.
• There is a need for a versatile model that can handle diverse correlated RNA-Seq data structures and imbalances.

Purpose of the Study:

• To propose a novel statistical model for analyzing correlated RNA sequencing (RNA-Seq) data.
• To develop a framework capable of handling various experimental designs, including paired, clustered, and longitudinal studies.
• To accommodate unbalanced datasets within the proposed modeling approach.

Main Methods:

• The proposed model assumes exon counts follow a multivariate negative-binomial distribution.
• It incorporates a cluster-level normally distributed random effect to account for correlations between samples within pairs or clusters.
• The model provides explicit expressions for marginal correlations at different levels.

Main Results:

• The model's performance was validated through a simulation study.
• It was successfully applied to two real-life RNA-Seq datasets: a paired study of clear-cell renal-cell carcinoma and a longitudinal study of Lyme disease.
• The framework effectively handles correlated and unbalanced RNA-Seq data.

Conclusions:

• The developed model offers a flexible and robust solution for analyzing correlated RNA-Seq data across various experimental designs.
• This approach enhances the analysis of complex biological data, including paired and longitudinal studies, even with unbalanced sample sizes.
• The model provides a valuable tool for researchers working with diverse RNA-Seq experimental setups.