Imagine that I have four groups: Control, Disease, TreatA + Disease, and TreatB + Disease. My goal is to determine whether TreatA or TreatB can reverse the disease-associated transcriptional changes.

I have been told that the appropriate limma contrasts are:

TreatA + Disease vs Disease
TreatB + Disease vs Disease

and that the significantly different genes in these contrasts represent genes affected by the treatment.

However, I am struggling with the interpretation. For example, suppose GeneX has the following expression levels:

Control = 3
Disease = 5
TreatA + Disease = 5
TreatB + Disease = 10

My confusion comes from how to interpret these treatment-responsive genes in the context of disease reversal. Using the example above, GeneX increases from 3 in Control to 5 in Disease. Under TreatA + Disease, it remains at 5, whereas under TreatB + Disease it increases further to 10.

In this scenario, TreatA vs Disease would not be significant, while TreatB vs Disease would likely identify GeneX as a treatment-responsive gene. However, intuitively, TreatA appears to better prevent further progression of the disease-associated change, whereas TreatB seems to push the gene even further away from the control state.

This makes me wonder whether genes identified in Treat vs Disease contrasts should necessarily be considered the most biologically relevant when the objective is to assess disease attenuation or reversal. Could it be that genes showing little or no difference between Treatment + Disease and Disease are actually reflecting successful stabilization of disease-associated expression changes? Am I misunderstanding the purpose of these contrasts, or is there a distinction between identifying treatment-responsive genes and identifying disease-reversing genes?