Hello;

I am running an 50 gene assay using targeted RNA-seq to verify results from microarray data. Thus, all of the genes are expected to be differentially expressed between samples from two conditions, and will have a lower range of expression than one would normally see in whole transcriptome sequencing. We can of course include genes that are not differentially expressed as well, although I am not clear on how many (if any), would be necessary to normalize for library size. I am in the process of setting up the analysis workflow and was wondering if DESEQ2 would be appropriate?

Many thanks for any suggestions.

Seanna

You can use DESeq2, but you'll need to come up with a way to normalize for library size, either using a set of genes that are not expected to be differentially expressed or some other external controls. If you just provide DESeq2 with the set of 50 known DE genes, then it is likely that the DE signal will be absorbed by the size factor normalization (depends on the distribution of up/down fold changes), or in any case such an input cannot be trusted to give reasonable results.

You can use estimateSizeFactors() before running DESeq() and specify controlGenes.

Or you can estimate size factors using some custom method and supply these via sizeFactors(dds) <- x, before running DESeq().

With so few rows, I would suggest fitType="mean" when running DESeq() rather than the other fit types, as it is a simpler dispersion~mean relationship to fit.