different output after changing group parameter

[mikagm]

Hi! I have a very different output every time I try to specify the biological variable using the “group” parameter of ComBat. We have RNA sequencing results from two different sites. I have 3 different groups and two different ages. I got first RNA-seq results from one age and then we increased the number of replicates and included another age but using a different sequencing company. I used ComBat to corrected the batch effect of the samples from the 'young' age and that resulted in 508 DEGs using DESeq2. Then if I want to compare the 2 different ages I need to correct the batch effect of all my samples. I used group=age and the results decreased to 181 DEGs. If I take the group into consideration using covar_mod the results decreased even more (98 DEGs). If I only remove the batch effect with group=NULL that gave me 349 DEGs.

I would appreciate any input/thoughts on this.