This is Going to Be Big (I Think): A Look at Numerate

In recent weeks, I’ve heard a few VCs talk about the promise of “biocomputational” startups, biotechs that are more quickly than ever collecting, processing, and making use of genomic data. Gene Security Network, which produces a test used to screen chromosomes during the in-vitro fertilization process is one that’s raised $10 million in the last couple of years, including from Sequoia Capital, Claremont Creek Ventures, and Alafi Capital.

Another that looks to hold potentially even more promise is a drug-design technology firm called Numerate, whose technology — from what I gather — is far enough along that it will be exceedingly hard for competitors to catch up. That’s largely because the company been tinkering with that technology for seven years; it emerged from a failed version of its earlier self called Pharmix, a biotecth that burned through $20 million of capital from Mohr Davidow Ventures among other investors.

Some but not all — including MDV — stayed for the ride at Numerate, which has already raised $9.5 million over two funding rounds since November. They may finally see that commitment pay off, too. Last month, the company was selected to negotiate what may be up to a $6.5 million grant from the Defense Threat Reduction Agency, a unit of the Department of Defense. It’s already received a much smaller grant from the National Institutes of Health. And Numerate is in discussions with nearly every big pharmaceutical company in the world, four (that we can’t disclose publicly) with which it is already collaborating.
So what does Numerate do, and how? Earlier this week, I spoke about the company with CEO Guido Lanza, who’d cofounded and served as CTO of Pharmix. I also rang Adam Grosser, a managing partner at Foundation Capital, which has become Numerate’s leading investor despite that Numerate is so far from the Foundation’s typical focus areas that there’s no place for it —  yet — on the venture firm’s Website.

Adam, why Numerate?

AG: For starters, the drug discovery process — particularly the R&D process — is broken. The FDA has approved about 20 drugs per year from 1980 through 2006. Meanwhile, the R&D expenditures required for 20 new drugs in 1980 were $3 billion and by 2006, they were $41 billion. The process isn’t working. It’s not viable to spend $2 billion to get a drug through the process.

Why has the process grown so expensive?

AG: Largely because of the human decision-making and intuition involved. You have to identify a target, then have a sense for what might bind to it and have the effect of expressing or inhibiting something, and you have to be right. And that’s just the very beginning of the process.

In fact, the process takes so long that intellectual property laws are having to be modified. I was talking with a scientist at Stanford recently who began work on a target 16 years ago and the drug will get approved by the FDA this year. Because patents are good for 17 years and most patents are patented at the beginning, the [U.S. Patent and Trademark Office] has had to come up with this crazy process of extensions; otherwise, no one could capitalize on their own work.

So how is Numerate trying to speed up the process?

AG: It’s basically narrowed down the process of selecting which drug is likely to work from a very large set of drugs. Its technology can quickly discern what makes a small molecule toxic, how you deal with bias and noise and data from a tech standpoint.

Put another way, it’s taking the human decision-making piece and transforming it into a data-driven approach where you can take billions of ideas and evaluate them. In the end, it has solved the computer science problem of drug discovery.

But isn’t there also a bottleneck issue at the FDA that needs to be overcome?

I don’t think it’s the FDA’s fault. They’re necessarily cautious. You don’t want to put things in people that are risky. But if you could take five to seven years out of the front end, you could get to clinical trials much more quickly.

Guido, how does the business side of things work, and how well-protected is your company?

GL: Actually, most of our information we’re able to keep as trade secrets because we don’t sell the technology. We partner on the output. A pharmaceutical company will come to us with an idea — maybe they are looking for an anti-cancer drug that hits the following targets but not these other targets yet is localized to the brain. That might be a specification we’d start with. From there, we go and design a compound using this technology they’ve never had access to. If it performs as we say it will, we receive eventual royalties on the sale of the drug. That’s the coin of the realm in our industry: compound intellectual property.

What can you say publicly about your technology?

We’re basically shifting the trial and error piece to the computer, and running that whole process on the cloud, through Amazon. We have access to as many computers as we want at any given time. If we need 500 to 1,000 computers for a month to tell us which compound is worth making, we can do that, and we can easily do five to 10 projects a year as a result. The next generation of our technology, which we’re working on now, will be even more of an accelerator for us.

How many compounds can most other biotech companies produce in a year’s time?

GL: Small biotechs can work on two or three maybe, while large ones like Genentech may work on tens of projects.

So what’s happening in that time that you’re producing a compound for a customer?

GL: Based on the specifications we’re given, our chemists then begin to look at a focused area of about a hundred billion compounds, which takes about a month. Then we start to look at all these ideas and our [technology] begins to rank them by likelihood of success. Over the next couple of months, working around the clock, the computer is combining different ideas and it’s getting smarter and smarter as it goes. Then you have to go to the lab and make the compounds and that takes a couple more months.

Why will Numerate work where Pharmix failed?

GL: Pharmix, which developed the previous generation of technology that we’re using now, had the wrong business model. We thought then that if you have something that generates small molecule drugs, why  not use it yourself? But there were reasons why that wasn’t the right model, in terms of which risks we could take on and mitigate. Other companies are better at picking out the next cancer drug.

Who are Numerate’s most direct competitors?

GL: There are really two in question: people who outlicense compounds — traditional biotechs, for example, and people who will run the traditional process for you, often in India or China. We have lost a deal when a company trying to do a drug design has gone out and licensed a compound that’s farther along, using a firm [abroad] because it can be much cheaper. The problem is that you don’t necessarily wind up with something you can use. There’s a design and not just an execution piece that [pharmaceutical] companies should be focusing on.