Dear all,
I am using DESeq2. My design matrix have two columns: status(1,2), and condition (C1,C2). I am only interest in detecting what genes are D.E. between the status. To do so, I wanted to use the likelihood ratio test (LRT). And to explore how the test is working, I created two dds :
dds1 - additive (design = ~ condition + status)
dds2 - interaction (design = ~ condition + status + condition:status)
Then, I respectively ran two LRT tests :
LRT1 - (dds1, full= ~ condition + status, reduced= ~ condition )
LRT2 - (dds2, full= ~ condition + status + condition:status, reduced= ~ condition )
The number of significant genes with FDR<0.05 were 529 and 260, respectively. Moreover, the significant genes within each test did not overlap entirely (only 200 genes overlap). Because LRT2 test whether there is any effect of status, condition-specific or not; I don't understand why I found less genes than LRT1
Can I rely on results from LRT2 ?