Super-delta2: an enhanced differential expression analysis procedure for multi-group comparisons of RNA-seq data

Area of Science:

• Bioinformatics
• Computational Biology
• Statistical Genomics

Background:

• RNA sequencing (RNA-seq) data analysis requires robust methods for multi-group comparisons.
• Technical noise and biases can impact the accuracy of differential gene expression detection.
• Existing methods may face challenges with statistical power and type I error control.

Purpose of the Study:

• To develop and evaluate super-delta2, a novel pipeline for differential gene expression analysis in multi-group RNA-seq data.
• To improve statistical power and control type I error rates compared to existing methods.
• To provide a computationally efficient alternative for RNA-seq data analysis.

Main Methods:

• Development of super-delta2, incorporating a customized one-way ANOVA F-test and post-hoc tests.
• Implementation of a multivariate normalization procedure with trimming and bias-correction.
• Utilizing large sample theory based on the Negative Binomial Poisson (NBP) distribution for asymptotic applicability to log-transformed read counts.

Main Results:

• Super-delta2 demonstrated superior statistical power and controlled type I error at the nominal level across simulation settings.
• Comparison with limma/voom, edgeR, and DESeq2 showed super-delta2's enhanced performance.
• Application to a breast cancer dataset identified more biologically relevant enriched pathways associated with pathologic stages.

Conclusions:

• Super-delta2 achieves tight control of type I error through trimming and bias-correction in normalization.
• The method leverages asymptotic normal approximation of the NBP distribution, avoiding computationally intensive iterative procedures.
• Super-delta2 offers an efficient and reliable tool for multi-group differential gene expression analysis in RNA-seq data.