Comparison of alternative approaches for analysing multi-level RNA-seq data

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• RNA sequencing (RNA-seq) is a key technology for quantifying RNA and understanding gene expression patterns.
• Accurate gene quantification and differential expression (DE) prediction are crucial, but challenging due to data noise and biases.
• Subtle gene expression differences can be difficult to detect with existing RNA-seq analysis methods.

Purpose of the Study:

• To investigate and enhance quality checking for mRNA sequencing (mRNA-seq) data.
• To develop and evaluate novel methods for RNA-seq data analysis, particularly for nested experimental designs.
• To improve the reliable prediction of subtle differential gene expression.

Main Methods:

• Incorporated sample layout information into analyses to accommodate multi-level experimental designs.
• Employed subsampling without replacement-based normalization for improved data comparability.
• Developed a DE identification method accounting for experimental hierarchy and effect size amplitude.

Main Results:

• The dataset-tailored methods significantly improved sample comparability.
• Achieved robust prediction of subtle gene expression changes in both D. melanogaster and H. sapiens datasets.
• Demonstrated enhanced performance compared to existing RNA-seq analysis approaches.

Conclusions:

• The proposed approaches offer potential improvements for key steps in RNA-seq data analysis.
• Incorporating experimental structure and characteristics leads to more robust gene expression predictions.
• These methods can enhance the reliability of findings in environmental, biological, and medical research using RNA-seq.