I am interested performing time-series analysis of RNA-Seq data. There are are about 15 time points, and the expression at each time-point needs to be compared to the baseline.
However, I am only interested in a small number of genes, around 6. Imho, it does not make too much sense to perform a DESeq2/limma analysis of all the genes, since the statistical hypothesis I am testing is much more moderate. I need to adjust (for multiple comparisons) only the tests on the genes of interest, not on all the genes.
Plotting these genes is necessary. But beyond plotting, is there some way to ask for the expression at each time point: whether the change in it (compared to the baseline) is statistically significant?
I think you are going to end up doing a full DESeq analysis anyway. You have all that data, why not use it? Surely a handful of rtPCR would have been better if you really only card about 6 genes.
The way to ask if the changes are significant is DESeq. The reason is that unless you have scads of replicates, you need the algorithmic tricks of something like DESeq to squeeze all the power you can get out of your data.
DESeq adjusts the p-values per each comparison. Since I am interested in only 6 genes, I do not need to adjust for all genes per comparison, but I do need to adjust for all comparisons being made.
Do you think I could take all the 6*14 p-values of DESeq (6 genes, 14 time points compared to the base line) and simply apply a multiple-comparisons adjustment for all of these ?
Or would you adjust for multiple-comparisons in another way?
Log in to answer this question.
