I can't believe they've received three quarters of a billion dollars of investment.
Given the slowing pace of drug development, that was probably a very good "investment" by GSK. But maybe it was a bad one for 23andMe if it limits the scope of deals they can do with other pharmaceutical companies.
A researcher friend of mine made fun of me for signing up to 23andMe when they first offered their test. SNPs, he told me, didn't tell you much of substance. To which I replied, haven't the last 20 years (almost 30, now) of genetic research been built almost entirely on SNP-based studies? He didn't have a comeback for that--he was in the middle of his Ph.D at the time. A few years later he signed up for 23andMe =)
It's true: we're so incredibly far from the type of personalized medicine that genetic researchers promised. I'm not sure what "transformative" would look like, but I don't think 23andMe's valuation exceeds its potential, if only because of the number of 0's in medical expenditures, globally but especially in the U.S.
SNP analysis is useful, but leaves out a lot of important variation. Also, it's taken the field a long time to understand that most interesting phenotypes are complex nonlinear functions of many different variations, which makes health discovery far more challenging.
That's true of everything in medicine, especially without SNP analysis. And it'll remain true even with whole genome sequencing.
I think the biggest problem is that until now medicine has revolved around the study and treatment cohorts. Even once we identify the variations, we still need to develop particularized treatments. It requires a tremendous shift in how medicine has been practiced. I think that's what 23andMe was hoping to bootstrap, and why they're not shy about sharing their database with researchers.
Moving from traditional medicine to individual treatment will be done incrementally--splitting what was once a monolithic group into ever smaller subgroups until, eventually, centuries from now, everyone receives tailored treatment. SNPs can still be useful because we're still transitioning from treating monolithic groups, so the variations missed by SNPs are details that we would be unable to leverage, anyhow. The state-of-the-art is always light years ahead of what we can manage to operationalize at scale. Isn't this the lesson so many startups have to learn? On the ground, the real value add always comes in figuring out to make some new technique useful and profitable, and for various reasons it's usually the case that the techniques are already ancient by the time they see substantial use in the market.
Finding better candidates for clinical trials, and then running better clinical trials are the two biggest things that have a positive impact.
There ya go.
23andMe was a vanity project for Sergey's wife (now ex-wife.)
https://en.wikipedia.org/wiki/Anne_Wojcicki
Same story for Marissa Mayer.
It's a caution to the rest of us, though. My annual salary is just a rounding error to them, and it's not worth their time to consider my individual fate in any way.
This still isn’t as bad as Elon Musk using Tesla to bail out his cousin’s failing Solar City.