This post was flagged by the community and is temporarily hidden.
You’re correct in your interpretation of the plots. When there are features over-represented in one level of the class only, the default is to display those in the positive direction. When some features are over-represented in one level and others over-represented in another (as you see in your first plot), the choice of which group to display in the positive direction (and which colors to assign) is chosen by the alphabetical order of the labels (“control” and “obese” in your case). If you changed the labels so that the obese label came first alphanumerically, (“A.obese” and “B.control”, for instance), the order would change.
I hope that helps!