Since DEXSeq::estimateDispersions() and DESeq2::estimateDispersions() are often the most time-consuming step of the two pipelines I was wondering if it was possible to share the result of this between a (naturally paired) DTE and DTU analysis using DESeq2 and DEXSeq respectively or if they are indeed different?

Cheers

Kristoffer

hi Kristoffer,

They are indeed different because the DTU model additionally shares coefficients and therefore conditional expected values of counts within gene, whereas DTE looks at each transcript on its own.

An idea for speeding up these models is to see to what extent you can use fitType="glmGamPoi". I think it should work out of the box for DTE, but may need some backend work for DTU. glmGamPoi should work well when you have a moderate to high number of samples, whereas the standard DESeq2 dispersion fitting is most important for low to moderate sample size.