Howdy, bhickey. I of course remain concerned about the following:
(1) On a population scale, common variants of modest effect are the most informative. These same variants are genotyped by 23andme.
However;
(2) On an individual basis, rare variants of large effect are the most important. By and large, 23andme does not ascertain these.
So people are missing many variants that are important to them individually; also, the interpretation of most known variants is by inference or prediction, not experimentation.
Let's not forget that, at this point, clinical risk factors still outperform genetic risk factors. Adding genetics to your clinical predictors, for most diseases, does virtually nothing for you. And why should it? Clinically, we can detect that you have high LDL. The fact that your genetics also predicts high LDL is irrelevant unless it can offer qualitatively different information. Risk prediction for the non-prenatal realm is not a particularly interesting use of genetics. For discovery of therapeutic targets, on the other hand, this is golden.
Genetics is still hard. Personalized genomics, much harder still. I'm not opposed to people getting access to their own genetic information, but I support very cautious interpretation. In medicine, each test has a risk: not only the risk of the test itself (e.g., radiation or bleeding), but also the risk of triggering follow-up procedures.
