Benchmarking differential expression analysis tools for RNA-Seq: normalization-based vs. log-ratio...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Next-generation sequencing count data requires normalization due to arbitrary library sizes.
• Log-ratio transformation is an alternative to normalization for analyzing compositional data, such as in 16S rRNA studies.
• Benchmarking transformation-based tools for RNA-sequencing data analysis is crucial.

Purpose of the Study:

• To evaluate the ALDEx2 package, a transformation-based tool, for detecting differential expression in RNA-sequencing (RNA-Seq) data.
• To compare ALDEx2's performance against conventional RNA-Seq differential expression methods like edgeR and DESeq2.
• To assess the utility of log-ratio transformations for RNA-Seq data analysis.

Main Methods:

• Applied the ALDEx2 package to two simulated and two real RNA-Seq datasets.
• Compared ALDEx2 performance with established RNA-Seq differential expression tools.
• Utilized a previously benchmarked dataset to directly compare methods.
• Introduced and tested a novel iterative log-ratio transformation within ALDEx2.

Main Results:

• ALDEx2 demonstrated high precision and, with sufficient samples, high recall in identifying differentially expressed genes from RNA-Seq data.
• ALDEx2 showed consistent high precision across different log-ratio transformations.
• Performance was robust regardless of the alignment and quantification procedure used.
• The novel iterative log-ratio transformation improved simulation performance.

Conclusions:

• Log-ratio transformation-based methods, like ALDEx2, are viable for differential expression analysis in RNA-Seq data under specific assumptions.
• ALDEx2 offers high precision and performs well on real-world data, complementing its use in 16S rRNA analysis.
• ALDEx2 can serve as a unified tool for analyzing data from multiple sequencing modalities.