Sept. 12, 2007 | For at least a decade, systems biology (SB) has struggled to establish itself as something important. It was nudged into existence by powerful new instruments spewing out a flood of data and advancing computational power to interpret the data. Today, systems biology is entering an awkward adolescence during which much of its value will be revealed, but remain far from fully utilized.
The label -- systems biology -- is pretty awful, except, of course, for all the other even more awful labels that have been tried. More important than what it’s called is what systems biology seeks to do: transform biology and healthcare into a rigorous, predictive science whose fruits, it is hoped, will be a deeper, more-detailed understanding of biology and a vastly improved approach to drug development and the practice of medicine. SB would build on the molecular biology revolution and elucidate the wiring diagrams (and their rules) buried in the data.
Institute for Systems Biology founder Lee Hood’s P4 Medicine vision -- predictive, personalized, preventive, and participatory -- is perhaps the most ambitious framing of what systems biology may help achieve. Put more prosaically, SB eschews focusing narrowly on individual ligand-target interactions, and emphasizes specifying entire biological networks (and eventually full systems) in mechanistic detail. In theory, researchers could identify the best intervention points, the optimum agent(s), definitive biomarkers, and avoid many toxic side effects. It should also be possible to capture these networks in simulate-able models to conduct “virtual” drug discovery and clinical trials. (See sidebar: What Is Systems Biology?)
Currently much of this promise is just that: promise.
Big Pharma still wrestles with how important SB may or may not be, and how to adopt it organizationally. The latter challenge turns out not to be trivial. A few true-believer startups have formed to exploit SB, and at least one – Merrimack Pharmaceuticals -- has compounds headed into trials. Circling the drug-makers is a diverse gaggle of SB technology providers, many struggling to scale up their fee-for-service and software licensing business models and wondering if they must become drug-makers to flourish long-term. Inside academia and government labs, SB is suddenly hot.
In other words, it’s a great time to be watching the systems biology drama, and if you have a strong sense of humor and sturdy constitution, probably a great time to be a player, too.
Not surprisingly, Big Pharma pays the lion’s share of SB bills, and it’s sending mixed signals. Several early SB advocates, notably Novartis and GlaxoSmithKline, dismantled or de-emphasized internal systems biology ‘departments.’ SB didn’t produce the hoped-for bang-for-the-buck, and thinning pipelines and the harsh financial realities plaguing all drug makers left little wiggle room for maintaining long-term technology bets.
“Systems biology at NIBR (Novartis Institutes of BioMedical Research) is no longer thought of as something distinct from the other disciplines which are brought to bear on our drug discovery process, such as biology, chemistry, and pharmacology. It is tightly integrated into all stages of the drug discovery process,” wrote Liam O’Connor, director of quantitative biology at NIBR, in a quote supplied by Novartis for this article.
That answer seems a trifle pat, though perhaps a worthy goal. Other observers suggest that Novartis – like many of its Big Pharma brethren -- is simply externalizing the risk through collaboration, and these observers hasten to add that’s not unreasonable. SB is young and unproven. Even Hood praises Novartis as forward-thinking and thinks it would make a good partner.
Contrarily, Pfizer, the largest pharmaceutical firm in the world, whose staggering R&D budget (roughly $7 billion) draws criticism and awe in equal measures, is not merely sticking with a distinct internal systems biology group, but also expanding it as part of a larger company-wide organization set in motion by CEO Jeffrey Kindler in January.
It should be noted the crux of Kindler’s plan calls for the ascendance of therapeutic area centers and, to a considerable extent, the descent of global technology centers. The company recently announced organizational changes to pour more money into finding and funding young companies working on drugs, not technologies. Pfizer may change its SB tune if costs mount and results do not.
Still, David de Graaf, Pfizer’s director of systems biology, is betting SB’s future inside Pfizer is robust. “Systems biology is now an organization that runs all the way into the early phases in the clinic rather than just providing candidate drugs,” he says. “It now needs to go through essentially proof of principle. [Management] decided that systems biology is an area of growth, and they’re going to set up a centralized group that will develop these approaches here at the Research Technology Center. The site mission is now being focused on deep knowledge of pathways, targets and compounds.”
The Pfizer group, based in Cambridge, Mass., is roughly two years old, and plans call for expanding it from eight to around 30 researchers. The wealth of local talent has made recruitment of the necessary multi-disciplinary expertise easier, de Graaf says, and Pfizer picked up at least one prominent former Novartis computational biologist: Carolyn Cho.
Early success in biologics projects and toxicity prediction, says de Graaf, helped drive SB’s acceptance and now expansion at Pfizer.
“This move is a chance for us to show more broadly what systems biology can do for Pfizer,” says de Graaf. We are beyond the tantalizing hints stage. The work we’ve done on hepatic injury, which we hope to publish very soon, helped demonstrate that. We don’t only want to show that we can sustain that and grow that incrementally. We want to go into some new areas and really change people’s perceptions. Ultimately we hope to start to grow small groups doing systems biology at all the sites.”
The Pfizer SB team will focus on bio-therapeutics, pathway investments that the company has made, and on predictive toxicology. “The classical model is to just make hefty investments in a pathway and have eight or 10 projects that are all essentially directed against the same biological mechanism,” de Graaf says. “We hope to not only [reduce the project] number but also to pick the ones that will be winners. That will eventually limit its application to particular therapeutic areas. Logical bets would be places like cancer and diabetes and maybe inflammation.”
Eli Lilly is also committed to retaining a robust internal SB effort. Earlier this year it announced a major expansion of its systems biology activities based in Singapore.
“You start with target ‘x,’ you don’t know the connection fully with target ‘x’ prime in humans, right, so animals to humans. So we test the hypothesis in rats and mice, asking the question whether it’ll be useful in humans,” says William Chin, Lilly’s vice president of discovery. “If we could just bridge that gulf a little bit better, if we could be more predictive, we’d be much better off. We think that looking at pathways and systems is a better way of doing it. The whole idea is that these systems and pathways are probably better conserved than any single molecular target.”
Chin reports Lilly’s five-year-old center has produced “a good number” of new markers that are proprietary. “I can say they are largely in cancer. We have, for instance, utilized some of this information for pharmacokinetics, pharmacodynamic modeling in a particular cancer program, which has allowed us, for instance, to help convince the regulatory agency, the FDA, that as we enter the clinic that we have chosen the right doses.”
Merck also has had early SB successes. Rosetta Inpharmatics (Merck subsidiary) researchers Eric E. Schadt and Pek Y. Lum published the upbeat review “Reverse engineering gene networks to identify key drivers of complex disease phenotypes” as part of a “Thematic review series: Systems Biology Approaches to Metabolic and Cardiovascular Disorders” in the Journal of Lipid Research (Vol. 47, 2006).
They wrote: “Networks reconstructed from these data provide a richer context in which to interpret associations between genes and disease. Therefore, these networks can lead to defining pathways underlying disease more objectively and to identifying biomarkers and more robust points for therapeutic intervention.”
Colin Hill, CEO and co-founder of SB technology provider and reverse-engineering specialist Gene Network Sciences, says, “They are bringing the power of SB and genomics to bear on major diseases such as metabolic syndrome with rather striking results. [Schadt and Lum] essentially claimed to have solved metabolic syndrome with SB and genomic techniques. This paper suggests a big win for Merck after five years of work.”
So why aren’t the floodgates opening?
Finding the Formula
One persistent challenge is organizing internal SB efforts to ensure the work actually gets done effectively, that SB is propagated, and that the work tackled is important enough to change attitudes. Should, for example, SB groups be centralized? Would it be better to build SB skill sets inside therapeutic group? If a company goes the centralized group route, should it act as service centers or do independent work?
“I don’t think anyone has the recipe for how to do it,” says Reudi Aebersold, another distinguished SB pioneer, co-founder of ISB, and currently the leader of Systems X.CH, a Switzerland-wide systems biology initiative. “I can say from my experience at the Institute for Systems Biology that we learned a lot of things. One of the key ingredients is that people are close by and communicate, and not just like in a Webex meeting, but that they actually run into each other, they sit together at the coffee table and really interact informally. That’s where the action is, especially if you have people from different fields that have not necessarily talked with each other before. (See sidebar: Systems Biology Is Hot in Academia and Government)
“Whether they are all part of the same department or a group or disease area, is not that relevant, provided that the people who are supposed to interact -- and that usually means that it is greatly facilitated if they are in the same building -- share their coffee tables, share a library, and share other places where people run into each other. That was an extremely striking experience at ISB that I’ve never experienced before or since, for that matter,” says Aebersold.
To some extent, the approach depends on your definition of SB and which of its many disciplines you are using. “My direct experience is mainly with modeling in our group,” says Jack Beusmans, principal scientist, pathway capabilities, AstraZeneca. “For the foreseeable future, I think it will be done within the specialized group. What I’ve seen in trying to transfer that thinking to super-users and biologists is that it hasn’t really been successful. It’s a particular expertise, and at this point it’s an art that’s only going to be successful for the next four or five years within specialized groups, specialized companies for that matter.”
Let’s also not forget, notes Beusmans, that Big Pharma’s investment in internal systems biology is tiny compared to what it is spending on companies with promising compounds. AstraZeneca spent $14 billion this year for MedImmune. That dwarfs spending on Beusmans’ group. “I don’t think people view in the very near future that systems biology, and certainly modeling, have any transformative potential. It’s incremental,” he says.
Not that modest bets or incremental progress are bad. Modeling is almost always involved in the target decision-making process at AstraZeneca, reports Beusmans, and his small group will grow slightly this year.
“We’re not really identifying targets. That’s very hard,” he says. “What we do is, given a system and some biology that we know about it -- for example angiogenesis and the receptors and growth factors that go with that -- given what we know about its regulation, what would be good points at which to intervene? These are the kinds of questions we talk about. Sometimes you can identify points of attack that are clearly not going to work in systems like that, and that’s valuable information. Sometimes we don’t know enough and can’t say anything.”
Prying SB success stories from close-mouthed pharma is a challenge. Few wish to elaborate on promising programs. What’s more, most early SB-related successes have been no-go decisions. While stopping problematic programs early saves money – sometimes big money -- and fulfills the “fail early and often” mantra – it doesn’t excite drug-makers the way identifying the next blockbuster would.
Alex Bangs, CTO and co-founder of bio-simulation specialist Entelos, once lamented, “The way we’ve been able to get short-term impact for people, for example, has been in cases where we’ve predicted that a target was not going to be effective, and the partners have — and this has happened with us multiple times — the partner has chosen not to pursue it.
“Then results have been published later by someone else that showed it not to be effective. Then everybody’s been doing a lot of high-fives about how much money they saved by not pursuing that. [And] we’ve had people say, “Well gee, we keep giving people negative news about things.” And the answer is “Well, a lot of stuff fails. So it’s not surprising that there are a lot of negative answers.”
Landing truly important work is another hurdle. SB departments must sell their expertise to therapeutic groups. Those groups are often suspicious for many reasons, not least of which is the reluctance to share glory or control and internal reward systems that actually financially penalize the therapeutic areas for such collaborations.
Pfizer’s de Graaf agrees this is a dicey problem. He says he is now working with portfolio managers in therapeutic areas to identify projects. These high-level decision makers worry more about moving important projects forward than preserving pet projects, turf wars, or who gets the credit. He stresses a healthy dose of diplomacy is always needed.
In the end, Hood probably has right perspective. “What will convince [pharma] are examples of drugs that came out of these approaches that were produced rapidly and economically. The only question is whether or not, after you get those examples, it’s too late for them to get in the ball game,” says Hood. “I think what’s going to happen is there are going to be a lot of younger companies who see that this is the way to do things, and they’ll be getting set up to do these kinds of things.
“In the past, what pharma has tended to do is merge with these companies, and it was easy in most cases because they merged to take the product and threw the company away. Whether they can merge with a company and retain its unique individuality is a question. I know of only one example (Roche/Genentech) when that’s been done and that was done superbly,” says Hood.
Systems Biology Startup
Enter Merrimack Pharmaceuticals, formed in six years ago by a multidisciplinary group of MIT and Harvard researchers and post-docs. From the outset, they sought to leverage systems biology to find drugs, specifically for cancer and autoimmune disease.
“We were actually part of a large DARPA-sponsored program designed to put engineers, mathematicians, and biologists in the same room to try to do biology differently. Merrimack is sort of spinout for what we were doing,” recalls Nielsen, vice president of research and a co-founder. He was doing a post-doc in Peter Sorger’s lab at MIT (Sorger is now at Harvard Medical School’s Department of Systems Biology). “There was a core of people, especially chemical engineers and biologists, who saw new avenues, not just to understand biology to change the way drug discovery done.”
That’s hardly a new story, but most “SB” spinouts opted to become tool companies. Merrimack bucked this trend. It also didn’t take VC money. Rather, over the course of five years it raised $145 million from other sources whom Nielsen declines to name. Headcount has grown to about 75. Making allowance for permissible marketing bravado, consider the company’s description of its ‘network biology’ approach posted on its website:
“…We believe that the key to substantially safer and more effective therapeutics is a better understanding of the complex biology that underlies disease. To that end, we have developed a proprietary technology platform that enables the high-throughput profiling of complex biological systems: Network Biology…Rather than focus on individual molecular components, Network Biology seeks to understand the system dynamics that govern protein networks - the functional set of proteins that regulate cellular decisions. More than a research tool, Network Biology is a set of technologies designed to enhance the entire drug discovery, development, and commercialization process.”
Merrimack’s models are mechanism-based, initially informed by the literature, but quickly enhanced by the company’s extensive (and expensive) wet-lab to generate large data sets.
“These are differential equation type models that describe molecular interactions by knowing the concentrations and rate constants for how proteins interact,” says Nielsen. “You can build up fairly large models of the system you’re working with, and importantly, you can use those models to simulate problems that are not necessarily in your data set.
“We’ve spent a lot of our focus on growth factors in cytokine signaling. We can ask questions like, how do you actually best inhibit a given pathway, and look at what are the most sensitive targets in the pathway. There are some really nice computational tools for doing that.”
Merrimack is focused exclusively on biologics. Its first candidate, MM-093, was in-licensed from McGill University and therefore represents less of a proof-point of its approach. MM-093 disease targets rheumatoid arthritis, psoriasis, uveitis, and multiple sclerosis. The rheumatoid arthritis project is in phase two trials.
Three other candidates, MM-101, MM-111, and MM-121, are homegrown and target cancer. MM-121 is expected to enter phase one this winter. “It was the first pathway we built,” says Nielsen, “and despite being very-well trodden, we found insights that we believe are important, and I think we’ve gotten a few years’ leap in terms of the thinking there and the antibody we’re bringing forward.”
Merrimack has modeled key signaling networks regulating the biology of solid tumors, including breast, ovarian, and colon cancers. Based on this research, it reports having developed several novel approaches to attack and shut down the growth signals that drive the growth of tumors. MM-111 and MM-121 are furthest along in these efforts. Companion diagnostics also figure prominently in Merrimack’s planning, particularly given the heterogeneity of cancer.
Neilsen says, “Network biology is giving us a new view on cancer that’s not necessarily defined by sort of growth physiology or even pathology. Whether the tumor comes from breast or ovarian tissue or colon, doesn’t always define the molecular pathology behind it, the molecular drivers.”
Unlike many in silico-mostly SB plays, Merrimack also depends heavily on its experimental biology and also intends to manufacture its product as well. Conversely, the SB tool company universe has tended to divide into in silico-mostly or wet-mostly companies.
“We’ve embraced a lot of multiplexing technologies that allow us to look at protein expression, phosphorylation in a sort of highly parallel fashion, very high throughput. These are actually insights we developed at MIT and, in part, some of what the company was founded on. We’ve also just adopted assays that are out there and tweaked them and made them high throughput so we use a lot robotics,” says Nielsen.
“Experimental biology is obviously much more expensive than literature mining and computational biology, so yeah, we invested heavily in method development, equipment, and people,” he says.
All this is enough to quicken Lee Hood’s pulse and prompt salivating by many pharma venture arms (one wonders if perhaps some pharma isn’t one of the investors). Nielsen insists Merrimack’s plan is to remain independent. He also offers an opinion why Big Pharma has been less successful in adopting SB.
“I think most pharmaceutical companies, at least on the surface, are embracing these technologies, often by establishing systems biology institutes or departments,” says Nielsen. “What we see the major hindrance being is that the reason this works well for us is through the integration of this approach into the whole of drug discovery, and in fact, bringing these tools into simulating problems in drug discovery before you create a drug.”
“A lot of where systems biology has been applied in the pharmaceutical industry has been in later stages; they say come in and look at our clinical trials, tell us who are the responders and who are not, which is a very different paradigm from sort of engineering things de novo, which is what we’re doing,” he says.
Merrimack is not the only SB startup, but its combination of wet lab and in silico expertise and focus on mechanistic pathways make it, perhaps, the purest SB play. Avalon Pharmaceuticals (to be covered in another issue) is another of the handful and it’s already had a successful IPO. Leveraging gene transcription technology and proprietary cell-based assay technology, Avalon’s lead candidate IMPDH (inosine monophosphate dehydrogenase) inhibitor, AVN944, is in Phase I trials for heme malignancies and expected to move into Phase II trials soon.
Still, in clinical trial is not the same as having a proven drug, and that’s the fuel that will truly ignite the systems biology fire. No drugs, no bonfire. But everyone seems to agree the inferno is coming, just not on when.
Technology advances and cost reduction are still needed, agree most folks. Perhaps more knowledge is needed too, though Nielson scoffs at this idea, “We hear a lot of those arguments. We talk to people who say we’re not ready to get away from empirical science, and, we don’t know enough biology to run simulations yet. Good for us if that’s the attitude out there. We’ve certainly seen tangible benefits in building even very small models of small systems and already getting insight that I believe goes beyond what other people have, and it’s certainly a different kind of insight than other people have.”
Douglas Lauffenburger, director of the Department of Biological Engineering, MIT, and an architect of many of the ideas being pursued by Merrimack, says flatly, “I am entirely confident about the bright future of systems biology, even if some pharmas bail out in the short run -- this is a long-term revolutionary approach to biology, so it is patently infeasible for it to “dissipate.”
Gustavo Stolovitzky, manager of functional genomics and systems biology at IBM’s T.J. Watson Research Center, says, “I think it’s fair to say that systems biology hasn’t delivered some big clear success story. In that sense, people haven’t jumped to it full-fledged. But I think everybody knows that in the long run, it will be one of the ways to go, if not the only way to go.”
“I’m really reluctant to make any time projections because I’ve seen too many of those where people say, well, five to seven years away and then five to seven years come and go and it doesn’t happen,” says Aebersold. “I don’t see one particular inflection. I think there are many inflections whenever someone comes up with a new technology that measures, let’s say, some contextual relationships or a software, an algorithm. It’s not an endpoint where, like in the genome project, where if you’re Bill Clinton you go and stand in front of the camera and say, ‘Now it’s been done,’ (even if it hasn’t) and have a press conference.”
In some ways, systems biology is both a philosophy and a methodology. The philosophy is pretty clear: Try to understand in fine mechanistic detail how living systems work as integrated entities, pushing ever forward from pathways to networks to whole cells and so forth until the entire organism(s) is (are) encompassed. That doesn’t sound so different from just plain biology, though a change in language is sometimes needed to catalyze actions. (See sidebar: View for the Systems Biology Supplier Trenches.)
The methodology is more about coordinated use of the diverse tools required to connect the dots that philosophy wishes to understand. In practical terms today, it’s primarily tools to interrogate molecular (and other biochemical) systems and computational power and math to make sense of the data.
In this latter sense, many companies have fastened onto this or that technology and championed their use of it as demonstrating their systems biology credentials. Sometimes it is; sometimes it isn’t. Often such claims have left observers confused about what really constitutes systems biology.
The next decade is likely to be systems biology’s journey through adolescence to adulthood. Perhaps, as Pfizer’s newest SB department member Carolyn Cho suggests, the clearest signal of adulthood will be the disappearance of the term “systems biology” as it merges into the mainstream of biology.