Got a note from Harvard/MIT's George Church re: my story on personal genomics and 23andMe (he's on the 23andMe scientific advisory board and is quoted in the story).

Nice job on the article. Many of my colleagues agree. However some (including me) are concerned that the message about “limitations” of linkage-based chips is not coming through (despite the basic forthrightness of the first three Personal Genomics companies). The limitations are significant for research -- and even more so for diagnostics. Common diseases are not necessarily caused by common DNA variants – but also by DNA variants individually rare but collectively common. Associations can be statistically significant in a research population but have unacceptably high false positives and negatives in the individuals of that population. A good example is one of the oldest success stories in personal genomics -- BRCA1 & 2 which Myriad Genetics offers. Could those gene tests be done with any of the 3 new personal genomics chip assays? Probably not, since the causative alleles are “new” (i.e. less than 100,000 years old) and numerous. The causative alleles are not in linkage disequilibrium with the common alleles on the chips and hence require actual sequencing at Myriad. Despite the various limitations, early adopters may be richly rewarded by their insider-view if/when personal genomics takes off like personal computers. Thanks, --George

I agree absolutely with George about the risk of hype here, that the perception could easily emerge that a limited genotype might be taken as somehow an exhaustive total sequence, and that there's a gap between the general science of associations and the utility of these studies for any one individual. In my story, I tried to state as much (though frankly and obviously I don't have the grasp of the science and technology here that Church has).

And I know that 23andMe, in particular, is wary of leading customers to believe that there's any diagnostic utility in their service - they quite vociferously insist (again, as I point out in the article) that their service is just a starting point, and that for actual diagnosis of risk or propensity for disease, you'll have to go to a company like Myriad.

It'll be interesting to see how this plays out - will the FDA or another regulator decide that, caveats from the company aside, that consumers are under mistaken impression about these services and thus in need of some kind of increased regulatory protection? For my part, I hope not - I think we need to let this play out some before instituting new regulations on a fledgling industry.

(ps - "many of my colleagues agree." hmmm, I wonder who hated it?)