I have four data collection, each collection has three datasets. The collections are data from long/short time treated/untreated cells. In untreated samples, I presume that the genes can change along with short/long time. I want to find out the different genes in long-treated cells vs short-treated cells, but get rid of the time effect(long-untreated vs short-untreated). I set in limma: factor1(time) with group1 long and group2 short; factor2(treatment) with group1 treated and group2 untreated. The contrast is “long-short”. is the result what I want? thanks.
Yes, that sounds about right.
In Galaxy, DeSEeq2 can accept one or several factors and Limma has the “contrast” option available.
Bioconductor vignettes/tutorials/help: https://support.bioconductor.org/
I have another understanding. The contrast “long-short” gives the result: different genes both in treated and untreated groups?
What I don’t understand is, what limma do in multiple factors between different groups?
The tool is generating metrics that represent the change in one condition with respect to the other along with “batch factors”. These fold change and significance metrics are context dependent, and each set of data will not have just one result since re-arranging the factors impacts which fold-change is reported.
The tutorials I linked to would be a good test set since the results are known. You could try experimenting with that data in a few different ways, and with different tools (or options) to see what is exactly produced. You might even want to do that anyway when exploring data as a type of sanity check.