I am interested in a comparing expression of gene_A vs gene_B in a homogenous population. If you don't have the number of samples to do a non-parametric test how would one do a parametric statistical comparison?

My thoughts are as follows:

1. Gene length normalisation is essential. TPM (transcript per million) or equivalent.
2. Inter-library normalization needed
3. Log transormation needed (to make it approx norm distributed)
4. Should probably be a paired test
5. Variance is most likely different in the different genes

Can one do a paired Welch test?

Do you have other ideas for how to do this?

In advance - thanks!