A solution to the problem of expression level variability?