Editor’s Note: Systems biology (SB) is once again a popular buzz phrase, though most companies actually doing it don’t call themselves SB companies. Buoyed by promising results from a recent collaboration with Pfizer on drug-induced vascular injury, Keith Elliston, CEO of modeling and biosimulation specialist Genstruct, provides an insider’s view of industry’s gradual uptake of systems biology in this conversation with John Russell, Bio-IT World’s executive editor.
JR: It seems like “systems biology” is again popping up in marketing campaigns and that the number of citations in peer-reviewed literature is growing. Are more drug companies becoming believers and dabblers?
Elliston: Well, the drug discovery industry is kind of slow to adopt new approaches because people have been burned. So I don’t expect to see a huge churn. I expect to see the classic Geoffrey Moore “Crossing the Chasm” kind of thing: You’ve got a small group of companies that are visionaries and early adopters; a large group that is more conservative mainstream; and then you’ve got a number of companies that are the laggards. What we’ve been seeing lately are a couple of the visionary early adopters like Pfizer getting more active. So those companies have started to do more and are leading the charge for others to do more.
JR: Will systems biology get across that chasm?
Elliston: I think that at some point everybody will adopt systems biology in some form. I think we’re a few years away from that. What we’re going to see over the next two years is deep integration in the visionary companies. You’ll also have some visionary people in the mainstream companies that give it a shot, but they’re not going to be able to put any real momentum behind it.
We’ve already had a couple of companies that we talked to a year and a half ago come back and say, “We went off and decided we were going to do systems biology on our own because you guys seemed kind of expensive. And we spent more money than we expected putting together software, training people, buying hardware, devoting time, and we’ve come up with nothing.” And I said, “We don’t look so expensive now, do we?”
JR: That certainly suggests growing interest, doesn’t it?
Elliston: Well, a lot of interest doesn’t always translate into a lot of business. I was at BIO and I saw a huge number of small companies focused totally on developing compounds. There are few technology platform companies anymore. Everyone’s got a compound and you’ve got the pharmaceutical industry, which has a dearth of compounds and is looking to license all sorts of compounds. Everybody sees compounds as the way out of the short-term crisis, but this is very short-sighted. The industry is going to have to get out of short-term crisis mode and try to fix itself with long-term strategies, if it’s going to be successful.
We’re seeing that we can show success with systems biology, not just that it helps you make better decisions faster once you have the drug, but it can help you find drugs more efficiently. Period. You might say better drugs faster, but [making] better decisions faster is only a little part of it. Failing early is still failing, not succeeding.
Look at the vascular injury project [with Pfizer]. If you can get a handle on this toxicity, you can identify why these PDE4 inhibitors cause vascular damage, and we can identify early biomarkers to do this, we can start basically engineering out of these compounds the toxicity. Or if we find it’s not relevant to man, we can develop these compounds directly in man. And that will enable a whole new class of drugs. That’s how you start getting the industry out of its doldrums.
JR: How difficult is it to “teach” systems biology to client researchers?
Elliston: I think if you’re going to try to help them understand systems biology and help them make use of the tools themselves, you’re going to have a very long road. It has to be incremental. The first step has to be familiarity. I think the brilliance of companies like Apple Computer is the approach they’ve taken, which is to wrap it up, and just get people something that they can use without having to learn all the technology behind it. When we started working with the safety sciences group at Pfizer they didn’t have much experience with systems biology. What we were able to do is spend four months with them, take their key data, model it, define it, and work with them to give them some actionable results without them having to learn how to use our platform, tools, and methodologies. We basically delivered the iPod of systems biology. It gives us a foundation to start “teaching” systems biology, if that is what the researchers want.
JR: What’s your view of the intellectual property generated on projects? Can you capture some of it?
Elliston: My belief is that if the outcome of what we do is producing better compounds more efficiently, then we should be able to reap some of the value of that benefit. If I was going to try and raise money today for a systems-based pharmaceutical company, I’d have to raise a couple hundred million dollars. My ability to successfully pull that off in today’s environment would be questionable. Could I do it in two years? I’m not sure.
So the approach you need to take is, again, an incremental one. Systems biology can be effective at defining mechanisms that we can use to deliver better compounds, to deliver less toxic compounds. Let’s work with big partners to help fund the building of that infrastructure because if they can increase their efficiencies, it will provide them with significant benefits. Smaller companies [are] not going to be willing to pay what the big companies are, but if you can help them get their compound through to market and you get a royalty on that, that’s great and you can capture value.
JR: Has anything been published on the Pfizer work?
Elliston: There’s not been anything published to date other than Pfizer presentations from various conferences. One of the key things we’ve found with our commercial partners is that they’re very interested in following up on our results and making sure they understand and sort through the intellectual property issues before they publish.
We haven’t done a lot of academic work, but we just completed a collaboration with Lee Hood’s group [Institute for Systems Biology], which had some really nice data. To be honest, we’ve used this internally as a training program for one of our new teams. Since about half of our work so far is in cancer, we wanted to try out one of the best-quality data sets that we’ve seen in prostate cancer. We’ve just presented our key findings to them, and we’re now protecting some of the discovered intellectual property. We expect to publish that paper at some point in the near future.
JR: What mergers or alliances do you think are likely in systems biology?
Elliston: You see a lot of cooperation among the different pathway houses these days. GeneGo is working with Ariadne, and we’re seeing more partnering in those spaces. We’ve done work with Ariadne and Jubilant, and use some of their tools and capabilities. We find it useful to bring in third-party technology where it helps supplement our core competencies.
When I look at what may happen in mergers and acquisitions, I think product companies need to be product [software] companies. So I don’t see mergers of product companies with collaboration-based companies like us. What I would see is more of a combination of capabilities to build a more fully integrated company. For example, today we don’t do any data generation. One of the decisions we made early on was that if we were going to focus on what we thought the key linchpin of systems biology, it was really going to be in trying to understand what all the data meant.
Since we don’t control data generation, enforcing issues of quality, capability, [and] design are based on our ability to influence our partners. Sometimes, we’re more successful than other times. I think at some point it makes economic sense to bring that in-house, but if you do that too early what you’ve done is accumulate an expense and impinge your ability to be successful in your core business.
JR: BG Medicine comes to mind as someone who’s doing data generation.
Elliston: I think data generation is important in terms of being able to produce the quality and quantity of data you need. One of the nice things is our approach is tolerant of less than perfect data. We’ve also been very successful in establishing quality controls on the data that we will accept for doing modeling. At some point what I would see happening is the development of a more fully integrated platform, but I don’t see it for us right now because our partners are skilled enough at generating data and you know they’ve got the biology in hand. Analyzing a group of samples isn’t going to get you where you need to go if you don’t have the right biology. I think Doug Lauffenburger published the four Ms of systems biology: measurement, mining, modeling, and manipulation. It really takes those four things to do it all. Doug’s group at MIT does all four of those themselves, and thus, they can control things from end to end.
Right now, we do the mining and modeling and our partners do the manipulation and measurement. But I could see the potential of building out a company that would bring all those things together. Then you have to ask yourself, if you can do all that well, do you want to do that in partnership or do you, in fact, want to go after the ultimate output: better compounds. I think where it’s going to go, if it’s successful, is to the point of producing better compounds.
JR: Does that say something positive for a company like Predix Pharmaceuticals?
Elliston: I think it certainly would. But you’ve got to be able to actually get it done and you’ve got to prove that you’ve got some ability. The latest numbers I’ve heard is that while small companies are making up about 70 percent of new IND filings, their failure rates are almost twice that of pharma in clinical trials. So, if you’re going to be successful, you’ve got to have good output; you have to have good numbers, and you have to have high success rates. The company that can do that first is going to be very successful.
JR: Genstruct has bet big on inference technology. Couldn’t others copy that approach pretty quickly?
Elliston: What I can tell you is taking that approach [inference engine] is a very hard place to go. If you’re going to take an approach of using causality and modeling based on empirical data, combining that with what you get out of an inference engine, there are a lot of pitfalls. I think people on the Semantic Web are talking about doing that as well. We started doing this in late 2001 – early 2002, and working with Pfizer since early 2003. We’ve got a lot of experience and think we’ve been able to push the technology through a number of key issues. I don’t think it’s something that a lot of people are going to able to easily reproduce in the near term.
JR: What’s the barrier to entry?
Elliston: I think the barrier to entry, number one, is key intellectual property that we been able to protect in this space. We’ve got a very good intellectual property position as first mover in this kind of approach.
JR: Is that first mover advantage or patent protection?
Elliston: It’s both. We’ve filed core patents on key elements of the technology, which we think provide very broad protection for us. They’ve been enabled because no one has really tried to take this approach in this industry prior, and even from an academic perspective. One of the key issues with mathematical approaches to modeling [in contrast to inference approaches] is there’s a lot of prior art. There was very little prior art in this field when Genstruct came into it to define these kind of approaches to understanding molecular biology and molecular systems and pathways.
JR: Do you see the pathway folks trying to move up into your space at all?
Elliston: That was one of the interesting things from your panel.* Those groups are all clearly in the world of trying to provide software to enable people to do things, but I did hear a few people trying to get to the [modeling] level. I think we’ve been out there talking about this for enough time and we’ve demonstrated enough success that they’re trying to get into this space in a different way.
I do see these [other pathway] software providers trying to think about how they can get closer to the kinds of things we do. But I look at the world and [it’s] a place where people want their technology to work like their iPod—plug and play. If you’re going to try to push the edge of technology, it doesn’t work that way. There are a couple of highly specialized groups, like David de Graaf’s [director of systems biology] group at Pfizer, with very skilled people, and they like to work with cutting edge technology, and they’re able to apply some of those tools in sophisticated ways. [But] I think for the most part, the rest of the consumers of the tools are running some data sets and looking at some pathways, and basically copying and pasting it into a report and giving it to somebody. I don’t think they’re being used at anywhere near the level of sophistication those tools can support, and this is the level that the vendors must support to continue to generate sales.
JR: Who are the companies that come to mind when you think of competitors?
Elliston: Entelos is a company that is doing systems modeling, though using a mathematical approach. Our approach really focuses on modeling molecular systems, and uses high-throughput genomic, proteomic, and metabonomic data. Entelos seems to be focused on building virtual patient models and modeling higher-order phenomena. I actually think our approaches are complementary.
JR: So who do you worry about?
Elliston: The companies that I’m keeping an eye on are some of the software providers. I look at them and say, number one, if they decide they are not going to be product companies and are going to try and do what we do, they would have a head start over people starting from scratch. Number two, if they figure out how to enable pharmaceutical companies to do systems biology in-house in a way that’s profound, then I’ve got to see them as competitive. And the ones I watch are GeneGo and Ariadne right now, because they seem to be the most innovative.
JR: Do you think of Compugen being in your space?
Elliston: Well, Compugen is an interesting company. I guess in some ways you’d have to try and put them there because they are using a lot of computational technology to try and figure out areas of biology. It is a company I keep my eyes on, but Compugen has had a lot of changes in management of late. They’ve certainly delivered value for customers, but I’m not sure if they’ve reaped the benefits of the value they’ve generated. It sounds like they’ve been trying to be a drug discovery outfit and then talking about different subsidiaries. I don’t know if they’re strategically organized in a way to be a systems biology company just yet.
One of the things I believe about business, especially for small companies, is you’ve got to do one thing well and grow from there. I was listening to a show on NPR [National Public Radio] this morning and they were talking about the success of Google. Why did Google succeed in search where Netscape and Microsoft and Yahoo haven’t? Because they concentrated on one thing and are doing it well. That’s a great lesson for companies like Compugen. They’ve got to figure out what the one thing is.
JR: Do you know Optimata at all?
Elliston: I’ve seen Optimata. They look to me like they are trying to be more like Entelos. I think what you’re going to find is that we’re at the level of molecules, and they’re at levels above that. I think these levels will converge, but today those are very different spaces.
JR: Would Gene Logic fit into that category?
Elliston: Gene Logic could in some ways. The one thing that’s most interesting about Gene Logic of late is what they are doing in drug repositioning. They are generating a lot of data, but I think if they don’t leverage it with system modeling approaches, they’re going to face the same problem with complexity that we are often asked to tackle for our partners. It’s really figuring out how to harness the power that’s in the data that is critical.
JR: How important are positive projects like your recent work with Pfizer in persuading the user community that systems biology is viable?
Elliston: I think it is very important to demonstrate success. The pharmaceutical industry is incremental in nature. You can’t say you’re going to revolutionize drug discovery and expect people to believe you. What we have seen is that most users first go out and buy commercial software from one of the key software providers; they play around with it a little bit and find that it adds value but not enough to justify a huge investment. So what’s going to drive investment is someone else saying: “Here is a success story and it’s changed the way I do business.” Then others will be willing to make to the kind of investments necessary to get to a point where systems biology can make an enormous impact.
JR: Is the necessary threshold of success being reached?
Elliston: That’s a great question. Overall we’re seeing some success. I think you have to look at some of what Entelos has done and say they’ve had some good successes in applying their above-the-cell-level modeling and their virtual patient model. They’ve been able to help people make clinical decisions. If you look at trying to understand molecular events, I think we’ve shown some real successes through the work we’ve done with Pfizer and other companies. I don’t think we’re there in terms of showing the full value [of the technology], but we’ve been able to generate the kind of successes you need to keep building positive momentum.
JR: What’s ahead for systems biology and for Genstruct?
Elliston: What’s ahead for systems biology is the next big technological push in the industry. But I think it’s going to be a little slower burn than ones we’ve seen in the past. We’ve been doing this for about four years now and we’re starting to see the inflection point, so you’ll see us doing a couple of aggressive moves in the coming year. We’ve been able to demonstrate success with partners, and now it’s the time to turn it into a commercial push and see where it can go.
* Bio-IT World’s Life Sciences Conference + Expo, April 2006