And it's extremely hard to sell biologists on the idea that your model just happens to capture sufficient signal to be useful or to accomplish more than a simple regression line on one or two known-to-be-causal factors. But I do think the importance of context has become inescapable and any therapeutic approach that overlooks it will fail to deliver needed efficacy.

Personally, I believe the day of telling the patient, "Sorry, but since none of the existing drugs will work for you, we can't help further" -- that era of medicine is past. Using deep drills through the maze of each patient's relevant / causal / contextual data, it's nigh time for us to dig deeper and do better.