Hi all,

I'm working with GTEx expression data looking at the effect of a quantitative variable on gene expression. GTEx provides already filtered and normalized RNA-seq data, the same that they used to perform eQTL analyses. They normalize between samples using TMM and also across samples using an inverse normal transform (INT). I normally use the limma+voom method applied on count data using normalization factors calculated with TMM. I’m not familiar with INT and I’m not sure if this transformation is good to prepare the data for the linear modelling done with limma.

My question is: can I use the limma-trend method on this data? Does it make sense to apply trend = TRUE in eBayes function to adjust for heteroscedasticity here?

I would appreciate your opinion.

Thank you very much,

Alba