Comparison of normalization and differential expression analyses using RNA-Seq data from 726 individual Drosophila...

Area of Science:

• Genomics
• Bioinformatics
• Statistical Genetics

Background:

• No universally accepted method exists for RNA-Seq read count data analysis.
• Gene expression differences across genotype, environment, and sex were quantified in 726 Drosophila individuals.
• The study aimed to identify the optimal analysis approach for detecting differential gene expression.

Purpose of the Study:

• Evaluate filtering strategies, normalization methods, and statistical approaches for RNA-Seq data.
• Identify the most critical factors for accurate differential gene expression analysis.
• Provide recommended analysis strategies for RNA-Seq read count data.

Main Methods:

• Evaluated three filtering strategies, eight normalization methods, and two statistical approaches.
• Performed differential gene expression analysis and statistical power analysis.
• Utilized RNA-Seq data from 726 individuals with eight biological replicates per condition.

Main Results:

• Normalization method, data distribution assumption, and biological replicates are critical for RNA-Seq analysis.
• Common normalization methods like Total Count and RPKM did not align data well.
• At least three biological replicates are necessary for sufficient statistical power to detect three-way interactions.

Conclusions:

• The DESeq normalization method combined with a negative binomial generalized linear model (using edgeR or DESeq software) is the recommended approach.
• Low-expressed genes were removed post-normalization and fitting to the negative binomial distribution.
• Recommended analysis strategies are provided for RNA-Seq read count data.