Google Pours “Incredible” Computing Power into Antibody Drug Discovery With Adimab
Google is the undisputed king of Internet search and advertising, but its second act as a company might be to invent a new computer model for efficiently discovering targeted antibody drugs.
“Google is committing incredible resources to it. Incredible resources,” says Tillman Gerngross, the founder and CEO of Lebanon, NH-based Adimab. “The infrastructure alone is in the millions of dollars of raw computational power.”
Gerngross won’t say exactly how much money and manpower Google (NASDAQ: GOOG) is putting into his startup, so it would be easy to dismiss this as chest-thumping from an overzealous biotech entrepreneur who’s just trying to raise cash. But Gerngross isn’t in that tight spot. He’s the Dartmouth professor who founded GlycoFi to make faster, cheaper antibody drugs in yeast, and sold the company to Merck for $400 million in 2006. His new company, Lebanon, NH-based Adimab, has struck deals in the past year to produce antibody drug candidates for Merck, Roche, Pfizer, and one other unnamed pharmaceutical giant. Those partnerships have given Adimab enough cash to run for the next 10 years, Gerngross says.
Google first made its interest in Adimab clear back in October. That’s when its corporate venture arm led an undisclosed financing that included Polaris Venture Partners, SV Life Sciences, OrbiMed Advisors, and Borealis Ventures. I spoke with Gerngross at length about the strategy behind this investment a couple weeks ago at the JP Morgan Healthcare Conference in San Francisco. We met one day before Adimab held a board meeting not far from the Googleplex in Mountain View, CA.
What interest does a computing giant like Google have in a little antibody company like Adimab? First, a little background. Adimab is positioning itself as one of the emerging discovery engines for making targeted antibody drugs which can zero in on specific targets on diseased cells, while sparing healthy ones. The market, born in the late 1990s, now generates about $25 billion in annual sales for targeted drugs like Roche’s rituximab (Rituxan) and trastuzumab (Herceptin).
Traditional antibody discovery is time-consuming and risky. Adimab has developed its advantage with a fast yeast-based model that can be used to synthesize hundreds of antibodies against a certain target in just eight weeks of work, compared with six to 18 months of labor with the traditional methods used in biotech labs around the world, Gerngross says. Once that work is done, the major drug companies still need to spend years of labor and hundreds of millions of dollars testing those drugs in animals and humans, determining which of these hundreds of candidates bind the best with the target and have the strongest effect against disease.
While Adimab represents a huge potential gain in efficiency with its faster, cheaper platform for discovering antibody drugs on the front end, Gerngross and Google are thinking about the next big revolutionary change at the early phase of the antibody discovery process.
That’s where computers enter the picture. Biologists currently have many well-characterized models for the structure of certain protein targets on cells, and the number of those available structures is growing, Gerngross says. What they don’t have is a computing model that can take the 3-D information on the target structure, and build on it, to say precisely where a certain Y-shaped antibody might bind to its target, for example. Once a biologist knows that, the next step is to ask whether the antibody can elicit a certain desired biological response, like cell death, or an immune system reaction. It’s an immense mathematical problem, and requires “formidable” computing infrastructure, Gerngross says.
When Gerngross first met with Google representatives on the Dartmouth campus, he wasn’t sure this was anything more than science fiction.
“In the past, that was a problem that was beyond computational means. But we think it’s actually not that far away,” Gerngross says.
Based on the commitment Google has shown since it invested in Adimab in October, Gerngross says it may now be possible to identify the optimal antibody for clinical trials entirely “in silico.” That would subtract a huge amount of time and effort from the notoriously lengthy, and risky, wet lab drug development process.
If Google’s computing power can actually achieve this lofty goal, a customer will still have to come to Adimab with a specific target in mind, and Adimab will still perform its usual 8-week process to synthesize a batch of antibodies that binds with a particular target. The key difference will be in speeding up what comes later, by giving the pharmaceutical customer a precise idea of exactly which of those 100 antibodies has the best shot as a drug.
“If you now can say, ‘OK, we have the structure, now we can design specific antibodies to hit a particular domain of that protein,’ that’s a capability no one has,” Gerngross says. “When we saw the opportunity to do this, we realized, if this works, we are out of business. If you can do what we just described, then it will be very serious competition. We want to be ahead of it.”
Gerngross said he hasn’t personally met with Google founders Larry Page or Sergey Brin to get a sense of their interest in this computing problem. But Bill Maris of Google Ventures has joined the Adimab board in connection with the latest financing, and “he has the ear” of Google board members, Gerngross says. Adimab also has an important man on the ground close to the Googleplex. Its senior director of computational biology, Max Vasquez, formerly of PDL Biopharma, lives in the San Francisco Bay Area and meets regularly with counterparts at Google, Gerngross says.
None of this is to say that “in-silico” derived antibodies will arrive anytime soon. Adimab still has other tough R&D projects on its plate, like making antibodies that can reliably hit hard targets like certain membrane-associated proteins, ion channels, and G-protein coupled receptors that weave in and out of cell surfaces.
But I took careful notes on what Gerngross said about this, because of his track record with GlycoFi and his outspoken criticism toward hype in the biotech industry. He told me that many of his pharma customers have been burned by other companies that overpromise and underdeliver. “The pharma companies think they’re buying a VW Bug, and then find out they got a tricycle.”
Despite those strong views, Gerngross sure sounds like he’s promising big things from in-silico antibody discovery. He just doesn’t have a specific timetable on when it will arrive.
“We think there’s a way of getting there, and it will be very powerful,” Gerngross says.