Re:Google Life Sciences Exodus
by danieltillett on Monday March 28, 2016 @05:43PM
(#11377656)
The real nail is that each individual is unique. There are not X number of disease cases out there, there are Y number of more or less closely related disease states. We might say that someone is suffering from the disease of high blood pressure, but the underlying causes and hence ideal treatment is unique to that individual. Add in idiosyncratic responses and the drug discovery problem is almost impossible.
I went from academia to industry to learn how to do what Andy said: make biotech copy tech. In many ways, I was extremely successful: I built Exacycle, which took Google's style of warehouse computing and made it available to a number of biology researchers. The scientists who were successful with it basically had to drop a lot of the preconceived notions about how to solve problems- they were accustomed to spending a huge amount of time setting up their simulations or drug docking (months or years of prep work before running sims for a year). Exacycle converted this to a rapid cycle where they could test an idea in a day on 600+Kcores/sec, make a few tweaks, and run again. This greatly sped up the productivity of scientists.
Ultimately, these ideas weren't well received in biotech until we published a bombshell paper on GPCRs that spoke to them in their language. Even then, it wasn't a product that I think any biotech would want to buy (this is the part where Andy was spot-on). I think Google/Verily paid attention to what we had to say and hopefully, they'll be able to succeed in ways that biotech does not.
At this point, though, I think the nail is in the coffin for rational drug design as a multibillion dollar investment for pharma. It's just not productive enough. Instead, I think pharma needs to solve the inverse problem: for every drug that is already approved for something, what diseases can it impact (the RDD problem is "for a given disease, find a drug that modulates the disease").