Hello!

I'm following the instructions shown in the limma user guide (17.4; page 113). The main objective of our project is to recover a list of DEG (from > 4 independents studies) and conduct a meta-analyzes (as a posterior step). However, some datasets have conditions that aren't the focus of the study (for example 300 samples = 100 A, 100 B, and 100 C; but our objective is to use only A and B conditions). In this case, we have to remove this kind of sample before the analyses (C), or can we load all the information (and conduct Background correction and normalize), but carrying the differential expression only on the samples related to the study?

Thank you very much!

In general, it is best to analyse each independent dataset with all the samples as it was originally analysed. There is no need to restrict only to A and B samples nor is there any advantage in doing so. Including all the samples often has advantages in terms of normalization, batch correction or variance estimation.

Having said that, if the number of samples is large in each group (n=100) and there are no covariates or batches then analysing only the A and B conditions alone will also work well. So, in the end, it is up to you.