Partially addressing 1 and 2: As someone who has had Lyme and writes software I can't imagine how we would even get 10% of the way there on #1, and #2 just sounds like a recipe for disaster. Imagine if Reddit took on patient diagnosis.

For lyme specifically, the blood tests are generally inconclusive, symptoms vary from person to person and are inconsistent, and even doctors who specialize in lyme seem to disagree on core details of the diseases' effects.

If consensus can't be reached even at that level, and if that level of complexity is present in many other ailments (which I assume it is), how would be able to build a beneficial piece of software around that information?