I have RNA-seq time-course data (3 TP) with control/case design I'd like to do TPM normalization for the dataset but not quite sure what's the best way to apply it. Would it be most wise to 1) TPM normalize each time point individually and after that 2) normalize the whole dataset in one step? If so, how one can in practise do this - as TPM normalizes per sample? How to use TPM to normalize a certain group?

Aiming to do this as we've used DESeq2 for the analysis but what want apply GEE statistics for this and see how the results compare. Manual data normalization is needed for the use of GEE.

Really appreciate the input if someone could give an answer to this.

You can use a transcript quantification method to generate TPM expression estimates, e.g. I have worked on the Salmon method that handles a number of technical biases while estimating TPM. You can import these into R with the tximport package and you will find the TPM values in txi$abundance.