Some approaches in use
Calculate estimates ? of effects, their SDs, their associated
t = ?/SD statistics, a modified t, t* = ?/SD*, where SD* is
suitably defined, and which includes Bayesian analogues.
Many choose using one or more of these.
In general these quantities will be affected by overall intensity,
average A, and so far there seems to be no simple, general
solution to the problem of systematically incorporating A.
Mixture models are not gene specific, nor is the use of anova.
Much more to do on this most basic problem.