I have determined a set of differential genes for a set of healthy vs cancerous patient samples. However, for the cancerous samples, there are two groupings, a more mild and a more severe group.

What would be the best way to determine, from our set of differential genes, which generally correlate (positively or negatively) with disease progression, from control -> mild -> severe, or vice versa. Average expression values for single-cells in each group and look for linear relationship?