Algorithm development - looking at the outliers - type 1 diabetes research

I'm going to circle back around to this, but thought I would link the article that was so intriguing for me:

http://diabetes.ucsf.edu/news/diabetes-center-researchers-contribute-‘most-comprehensive’-human-epigenome-roadmap

My initial thoughts is what happens if you look at the parents who had Type 1 diabetes and looked at the markers in their children who either did or did not get Type 1 diabetes?
If Marson is on to something, noticing in the junk DNA potentially some markers for signaling development or effect, would there potentially be an advantage to looking at the junk DNA or more specific groups, especially those that did not get Type 1 diabetes?