Benchmarking RNA-seq differential expression analysis methods using spike-in and simulation data.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• RNA sequencing (RNA-seq) is crucial for gene expression studies.
• Differential expression (DE) analysis methods are widely used but benchmarking results vary.
• Previous benchmarks using spike-in data may not fully represent biological variability.

Purpose of the Study:

• To compare the performance of 12 RNA-seq differential expression analysis methods.
• To evaluate methods using both RNA spike-in and simulation data under negative binomial (NB) models.
• To identify robust methods suitable for various simulation conditions.

Main Methods:

• Evaluated 12 DE analysis methods, including recent software variants.
• Utilized both RNA spike-in and simulated RNA-seq data for performance testing.
• Conducted extensive simulations to assess the impact of DE gene proportion, dispersion, and balance.

Main Results:

• Method performance differed significantly between spike-in and simulation benchmarks.
• DESeq2, edgeR.rb, voom.tmm, and voom.sw demonstrated robust performance across various simulation conditions.
• Performance was sensitive to the proportion, dispersion, and balance of DE genes in simulation data.

Conclusions:

• The choice of benchmark data (spike-in vs. simulation) critically impacts RNA-seq DE analysis method evaluation.
• Simulation data provides a more comprehensive assessment of method performance, reflecting biological variability.
• DESeq2 and specific voom configurations are recommended for reliable DE gene analysis under diverse conditions.