“High throughput” biological datasets we can view as containing many sets of responses.

A typical example is RNA-Seq gene expression.

• Perform RNA-Seq on a set of biological samples.
• Number of RNA molecules seen for each gene in each sample is counted.

• Linear model formula based on experiment design.
• Each gene can be viewed as a separate set of responses.

Key features

• A test is performed for each gene. Multiple testing adjustment is crucial!
• Residual standard deviation is similar between genes. “Empirical Bayes” shrinking of these standard deviations overcomes problems with small numbers of samples.

Use Bioconductor packages such as limma (fit many linear models) or edgeR (fit many GLMs).

From the authors of limma and edgeR at WEHI

• “RNA-seq analysis is easy as 1-2-3 with limma, Glimma and edgeR”

• “A guide to creating design matrices for gene expression experiments”

From Monash

• MGBP’s RNA-Seq end-to-end workshop

• Degust for web-based differential expression analysis