I am assigned to analyse RNA-Seq data from a study with a repeated measures design with three time points. From what I understand (after going through Q&As in this Bioconductor forum and reading several manuals), the limma package is able to incorporate the within-subjects correlation via the blocking argument and the duplicateCorrelation command. However, a statistician in my university advised me to use lme4 for loops instead to fit every gene one by one (after voom transformation & using sample-specific weights obtained from the command).
The latter approach took a LOT of time to run (few hours), while limma only took a few seconds. In addition, working within the limma workflow enables me to use the included commands such as camera and goana seamlessly.
What do you usually do when you analyse repeated measures RNA-Seq data? How many of you use custom-scripts with lme4/nlme to fit mixed model for each genes? How common is the practice anyway?
Thank you very much for your time.