SAN FRANCISCO, June 24 - As genomics struggles through biotech's recent hardships, "drug development" is quickly emerging as the most-favorable business for companies originally built for selling tools and technologies.
But on either side of this notion is a giant rock and a brutal hard place. Big pharma valuations have taken a hit on drying pipelines--a fact that has financially bruised one of genomics' biggest customers--while scrappy young biotechs, another major client base, are picking up the slack and presenting themselves as capable competitors.
Companies such as Exelixis, a self described "genomics-enabled drug-discovery" operation, on the other hand, occupy a tidy little corner between the behemoth pharmas, the nimble biotechs, and the hard-pressed tool shops: an 'un-pharma' firm that's close to the tools companies, is structurally lean, and is directly focused on getting drugs into the clinic.
Geoffrey Duyk, chief scientific officer of Exelixis, recently sat down with a GenomeWeb reporter to pinpoint the "mortal sin" of modern genomics and catalog the next steps to make drug discovery via genomics a reality.
GenomeWeb: How is genomics evolving?
Geoffrey Duyk: People talk about genomics as a scientific revolution. I would argue that genomics has been more of a cultural revolution in the sense that its biggest impact has been on operational/organizational issues within science. It's the first time biologists have really linked themselves to high-throughput processing technologies.
What has been the mortal sin of genomics, because the technology can be both seductive and corruptive, is confusing throughput and output. There has been so much focus on throughput, not enough focus on understanding the output of what's going on. And so I think the tools are must-haves but in and of themselves they are only pieces of the pie.
Given what you have is this large, long integrated process where there are many sorts of hand offs, the most likely consequence of speeding one component is to create a bottleneck somewhere else. So we have to focus more on how to balance the hand offs, how do you focus on the bottlenecks, how do you make more scalable the components of the process?
GW: Is that bioinformatics?
GD: I think it's an operations thing. Informatics plays a significant role, but informatics not in the sense of in silico biology, but informatics in the sense of laboratory information-management systems. You have to be able to monitor your process and capture the information. I think there's been a lot of emphasis inside Exelixis in using informatics tools or computer tools to measure and monitor how we do research, not necessarily direct what we do.
GW: In what other ways can you make R&D more efficient?
GD: If you look at most biotech companies when they start their research, they tend to focus on one component of the research process because they have some insight and they're going to make that better. The process of forward integration is moving to the left or right of that point, trying to say 'I have some advantage here, how can I link the other components of research?'
So it's sort of like a football game: If you don't have a good offensive line, no matter how good a quarterback you have you're just not going to be successful. So I think it's more of understanding you have to balance the whole team. [With] sequencing, if you improved your methods for making DNA so you can make a million preps per day but the throughput of your instruments are only 100,000 it doesn't really matter. It just accumulates. Or you buy more machines, and at some point the mismatch is so great you can't compensate.
What Exelixis did was become what I'll call a sophisticated end user of technology. Not so much developing as a tools company the new technologies, [but] survey the technologies, take those that we considered robust, think of things in the context of being ultimately disposable. So knowing that, just like computers, technology turns over every two to three years, and really pay attention to the hand offs.
I think the companies that have been successful have really focused on the integrated platform approach and really have managed problems of drug discovery more than any one particular component. The end goal isn't the technology, it's the product of the technology. And at the end of the day your investor doesn't care whether you in-license your compound, you develop as a consequence of voodoo or human sacrifice or a rational technology-driven process.
So for us, because there is a kind of suspension of disbelief to be a small company competing with a Merck, you have to look very carefully at what you do well, and where you cannot compete. Companies like ourselves I think do well by thinking about the process of operational organizational efficiencies, decision-making processes, intellectual capital--I think we can attract to biotech industry a different group of people that you necessarily get inside pharma - so I think there's a different kind of attitude sometimes fostered. We tend to be much more open minded about technology and earlier adapters. Those kinds of things work to your advantage.
GW: What technologies now look good? What are killer applications?
GD: I argue that in biotech there's been no equivalent of the router. It's not like everyone else is using bows and arrows and we're using a gun. I really think the technologies provide incremental changes. And so if you think about entrepreneurial companies as being a balance between a technical advantage and a sociological/operational advantage, right now I think execution is probably in the forefront over technology.
But if I look across at what we're doing, what are the things that I think have really changed, on the informatics side is just the cost curves on storage media and processing power have made a huge impact. I don't think they're big changes, they're incremental, a reduction in cost. The other thing that has had a big impact is the commoditization of previously difficult-to-obtain technologies for small companies.
GW: Such as?
GD: Analytics. So if you look at the cost of mass specs. For example, we have a very significant combinatorial chemistry effort which synthesizes upwards of two million compounds a year. That's only possible because you can buy analytics that allow you to look at each and every one of those compounds and actually verify that 'I want to synthesize A and I actually got A.' So I think the fact that the cost of mass specs and things like that have come down in range, and have gone up in throughput, that you can actually afford them, are really important.
High-throughput screening technologies. A decade or more ago you'd have to have a relatively strong in-house engineering department to build the kind of robotics and software. All that today is obtainable off the shelf. A lot of specialized assay technology, the same thing: it's available off the shelf.
Sequencing technology is a good example: It's all off the shelf. I think on the crystallography side, in chemistry we do a lot of structure-based drug design so that was a pretty specialized field a decade ago. Today there's more people in labs that do it. The software is more commercially available. That's really the theme, this commoditization of technology. If you understand what you want and what you need, you can put the pieces together in a relatively cost-efficient manner.
GW: So what do you need next? What exists, and what do you wish existed?
GD: On the proteomics side, I don't think the issues are so much the mass specs. I think the limitations are on separation technologies. Your ability to actually purify and identify proteins, I think that really has to get better to take full advantage of the mass specs we have today. It's more a biochemistry problem than the mass-spec side.
GW: Is new technology needed, or more expertise?
GD: The mass specs got better faster and the engineering is easier than actually working through how do you purify proteins. Working out proteins is hard compared to working in nucleic acids. Some of it is just a trend in what people study. I did my PhD in the 1980s, everyone flocked from biochemistry and physiology to molecular biology.
There's not a lot of indigenous expertise out there right now in protein biochemistry; you're beginning to see it re-emerge. But I think you need that, and people coming from material science. Most of the interesting advances in biotech tend to come at interfaces. In fact, I think the more you blur the distinction between biotech and high tech the better off you are in terms of thinking about new technologies.
GW: Such as?
GD: In genomics the most important thing is getting the cost of sequencing and genotyping down. You're not going to see a revolution in human genetics until the cost per genotype is probably on the order of one hundredth of a penny or thousandth of a penny per genotype because I think you have to be able to do a lot of experiments very cheaply to understand the best way to do it. The same with sequencing. It's still a hundred million dollars to sequence a human-size genome to some high quality. That's a moon shot. Technology has to become disposable before it really has an impact.
GW: How so?
GD: You have to not care to use it over and over again. Right now to sequence a genome is practically an act of Congress, as opposed to, 'Oh, I think it's really an interesting thing to do.'
GW: What's a technology that went from big budget to disposable?
GD: Crystallography is a good example. It's not tossable yet, but it's gone from something highly specialized to something you can just build yourself. Oligonucleotides; now you can buy them for pennies a base. You don't have to think about them, they're not even in the capital-equipment budget.
GW: And to make an impact the same should be true with sequencing?
GD: Sequencing has to be like a centrifuge. It has to get to the point where sequencing a genome you can do for a thousand or ten thousand or a hundred thousand dollars. I think it will take a decade to get to the next cost level.
GW: How does that bode for biotech?
GD: Biotech doesn't need that kind of sequencing. The scientific community [does], but not biotech. The issues in drug discovery have less to do with sequencing and more to do with the fact that we just need to focus more on the biology. We've identified the component parts list, the genes, the proteins. We're getting increasingly good at seeing how those things assemble into biochemical signaling pathways. The question is how does biochemistry translate into first cellular and then organismal physiology. That's complicated.