Power estimate for multiple timepoint expression study

Question

[coleman_jonathan]

I am planning a study examining genome-wide gene expression changes across three timepoints (baseline, following a treatment, and some time after a treatment).

Does anyone know of any statistical package or equation to estimate the effect sizes a longitudinal analysis of this would be powered to discover, given a known sample size and modelling the data as a mixed linear model incorporating the effect of time?

Answer 1

[Devon Ryan]

Normal power calculations would be for just one of the genes, so you end up needing to use another dataset as a model for how yours will behave and then determining power according to that. One popular tool for doing that with RNAseq is Scotty. These generally work better if you've already done a pilot experiment (as is true with all power analyses, since otherwise you're really just making numbers up).