|Home | About | Journals | Submit | Contact Us | Français|
So far, little is known about systems biology and its potential for changing how we diagnose and treat disease. That will change soon, say the systems experts, who advise payers to begin learning now about how it could make healthcare efficient.
If you think of the human genome as a parts list, then systems biology is all about how those parts interact in a healthy human, and exactly what goes wrong when disease happens.
Systems biologists often turn to metaphors when they compare the old reductionist approach, of focusing on a gene or a protein as the culprit in disease, with the new systems approach of parsing genes and proteins and their quantitative dynamics in disease-perturbed networks. After all, the human genome consists of about 25,000 genes, each of which is responsible for multiple proteins and all of which interact in so many mind-boggling combinations that it takes supercomputers to keep track of them for just one disease or organ system.
Take cancer, for example.
“The only way we really have a chance of making sense of it so we can treat cancer in a logical fashion rather than an empirical fashion — which is how it’s been done in the past 50 years — is to try to figure out how oncogenes and tumor suppressor genes fit together in some kind of a network,” explains Lewis Cantley, PhD. Deploying his favorite metaphor, “We need to understand the network, just as you understand the electrical wiring in your house. If something goes wrong, rather than pulling fuses randomly, you replace the specific fuse that went wrong. That’s the goal.”
Cantley is professor of systems biology at Harvard Medical School and directs the Cancer Center at Beth Israel Deaconess Medical Center in Boston. In 2008, his laboratory discovered the PI3 kinase (PI3K) network, a potential target for cancer therapy because it is essential in regulating glucose metabolism, which fuels the growth of most malignant cells. He also is a coauthor of a 2006 article in Nature Reviews in Genetics that describes how the PI3K network controls cancer.
“There’s been a lot of progress in the last 30 years fitting together what we used to call signaling pathways,” Cantley continues. “We know that these are all really networks that crosstalk to each other in very complicated ways.”
Untangling those complexities is where mathematical computer modeling comes in. Only supercomputers like the IBM Blue Gene can crunch the millions of data points and differential equations that represent the networks of interest both qualitatively and quantitatively. By modeling at that level, researchers can make their way through the complexity of gene expression, gene mutation, protein interaction, redundant pathways, and negative feedback loops to zero in on the network node or nodes that can be targeted with drugs or RNA interference to shut down a disease.
It’s an iterative process — developing an in silico model based on the literature, clinical data, and in vitro experiments with cell lines; running simulations with that virtual model; testing the predictions generated by those simulations in vivo (usually mice); refining the virtual model based on the in vivo experiments; and then repeating the cycle with more in silico simulation. If all goes well, the final in vivo experiments take place in human clinical trials.
“A lot of systems biologists are 99 percent computational,” Cantley says. “We do simulations, but they’re invariably followed by experiments to test the predictions. The value of modeling is that it tells you what experiment to do, and then it helps you interpret the results of that experiment so you can go back and design the next one.”
PI3K turns out to be a major genetic mutation in breast, ovarian, endometrial, colorectal, and prostate cancers and in glioblastoma. It’s a “drugable” enzyme, Cantley explains, and every pharmaceutical company in the country now has a PI3K program — 15 phase 1 clinical trials of PI3K inhibitors are now under way. Cantley leads the Targeting PI3K Pathway in Women’s Cancers Dream Team. The goal is to discover approaches that will predict which patients will respond positively to PI3K inhibitors.
Scare up some PhD-level biologists, computer engineers, computational biologists, control theory engineers, and physicists, and you, too, can make a nice living doing systems biology for big pharma.
Alex L. Bangs worked in robotics before joining colleagues Tom Paterson and Sam Holtzman, PhD, who, in the early 1990s, started to develop a disease model to help a client leverage a clinical database. Back then, there were no off-the-shelf products for building complex biomodels, so Bangs started writing the software. Today, he is chief technology officer, and, along with Paterson and Holtzman, a cofounder of Entelos, a privately held predictive biosimulation company in Foster City, Calif., that specializes in developing “dynamic large-scale computer models of human disease.”
Entelos has evaluated drug candidates for a variety of diseases with collaboration partners GSK, Merck, Pfizer, Bayer, Organon, Johnson & Johnson, and Eli Lilly. The in silico approach can save time and money, says Bangs, by identifying drug winners and losers early in terms of efficacy, safety, and potential toxicity. “If you’re building a model of type 2 diabetes, you keep adding to it as people look at new targets and new populations. The technology becomes not just a computer model, but also a knowledge repository. Everything we’ve learned about type 2 diabetes is in this technology.”
In 2008, Entelos entered into a cooperative research and development agreement with the U.S. Food and Drug Administration to learn more about drug-induced liver injury. The collaboration also includes the Hamner Institute for Drug Safety Sciences in North Carolina and several pharmaceutical companies. The idea is to identify drugs with the potential to cause liver injury before they hit the market. Liver effects that are discovered after a product launch may cause a drug to be withdrawn or to get hit with a black-box warning, which could seriously affect its potential in the marketplace. The cardiovascular safety and efficacy of a class of drugs is the focus of another Entelos collaboration with the FDA.
Gene Network Sciences (GNS), in Cambridge, Mass., was founded in 2000. The human genome had just been sequenced, and many smart people were convinced that flipping a few genetic switches could cure disease.
“The hype was a little premature,” says Thomas A. Neyarapally, JD, MBA, GNS’s senior vice president of corporate development. “We essentially had discovered the alphabet, and people expected that within a year we’d have Shakespeare.”
Systems biology may not yet be at Shakespeare’s level, but GNS is now putting together coherent stories about human biology for clients such as Johnson & Johnson, Pfizer, and Biogen Idec. GNS collaborations with academic centers are also in the works. An in-house talent pool assembled from the upper strata of the global pharmaceutical, biotech, and academic communities wields the proprietary GNS Reverse Engineering & Forward Simulation supercomputing platform to translate massive quantities of genetic, genomic, and clinical data into so-called biomodels of disease. In silico simulations with these biomodels will be able to identify the molecular drivers of disease and drug response.
“Most of the low-hanging fruit has been plucked,” says Neyarapally. “We now have to delve further to identify and understand the molecules that maybe aren’t so neatly and clearly correlated with the disease that we’re trying to understand but are very highly connected ‘key nodes’ in a network that enables the disease to occur.”
Discovering novel intervention points in a disease network is a good day’s work, but that same BioModel often suggests diagnostic, prognostic, or drug efficacy markers. Sometimes, the goal is to find a biomarker of drug efficacy, as in a recent GNS collaboration with Johnson & Johnson for an oncology drug. The GNS portfolio also includes collaborations on rheumatoid arthritis (RA), metabolic syndrome, and such central nervous system diseases as multiple sclerosis and Alzheimer’s disease. A collaboration with Biogen Idec focused on finding out why 30 percent of RA patients don’t respond to anti-TNF drugs. Longitudinal data from human subjects included measurements of certain proteins, gene expression, variations in single nucleotide polymorphisms, and endpoints like pain and joint swelling.
From Neyarapally’s perspective, the resources to do systems biology right have been converging in the last several years, so it may be another few years before a drug that can trace its pedigree to the systems approach hits the market.
“If you’ve discovered a new target, that’s great,” says Neyarapally. “But it doesn’t really matter until there’s a drug and that drug is in a patient, and that takes a few years.” There is an opportunity, he adds, to affect personalized medicine for drugs currently in development by biomarker discovery using systems approaches today.
No one has done more to marshal the resources for systems biology and put them to use than Leroy Hood, MD, PhD, cofounder and president of the not-for-profit Institute for Systems Biology (ISB), in Seattle. Hood has 12 patents, has co-founded nearly a dozen companies, and was awarded the 2002 Kyoto Prize in Advanced Technology for his contributions to biotechnology and medical technology.
“If you were an engineer, how would you figure out how a radio converts radio waves into sound waves?” is how Hood introduces his favorite metaphor for systems biology. “The first thing you do is take the radio apart and identify the component parts, and that’s exactly what the genome project did. They identified all genes, and by inference all proteins. Second, you would re-assemble the radio parts into networks and come to understand individually and collectively how those networks work to make radio-to-sound wave conversion. That’s exactly what you have to do in human biology.”
In retrospect, it’s easy to see how Hood’s career provided the insights as well as the key hardware essential to doing systems biology, as we know it today. For two decades, he was a molecular immunologist at CalTech, where Nobel laureate Max Delbrück’s comment that immunology is too complicated to study one gene or one protein at a time left a lasting impression on him. Hood also had been thinking about systems and began developing tools that made high-throughput data generation possible, including the automated DNA sequencer, the sine qua non of molecular and cell biology and the inkjet technology for making microarrays that was commercialized by Agilent. At about that time, the mid-1980s, the American scientific community was split 50/50 on sequencing the human genome. A strong proponent of the idea, Hood was influential in bringing his colleagues around to his way of thinking.
In 1992, Hood moved to the University of Washington to serve as the William Gates III Professor of Biomedical Science, where he created the first interdisciplinary department of biology, which included engineers, physicists, and computer scientists, and became the model for systems biology startups. In 2000, Hood cofounded the ISB with Alan Aderem, PhD, and Ruedi Ae-bersold, PhD. According to the ISB Web site, “The goal of the institute is to increase … understanding of the human immune system and other biological systems so that medicine can become more predictive, preventive, and personalized.” ISB partners include CalTech, Fred Hutchinson Cancer Research Center, University of Washington, Stanford, Rockefeller University, Institute for Molecular Cell Biology in Singapore, Pacific Northwest National Laboratory, Merck, and IBM.
The 300 ISB members have lots of irons in the fire, and Hood offers two examples of promising work. A complete analysis of messenger RNA in the progression of prion infection (Creutzfeldt-Jakob disease) in mice has identified key disease-perturbed networks that explain virtually the entire pathophysiology of the disease — potentially good news for mad cows and other critters at risk for prion infection, and maybe for people with Alzheimer’s disease. “We have looked at Alzheimer’s in humans, and we find that some of the mouse neuro-degenerative networks that operate in prion disease also are operating in Alzheimer’s disease,” says Hood.
In a related development, Hood reports that transcriptome analyses can identify transcripts, and hence proteins that are organ-specific and reflect network perturbations long before other clinical changes show up in the blood. The mouse brain, for example, synthesizes about 100 proteins that end up in the blood and constitute a molecular fingerprint of how the cognate networks for these proteins behave in the brain. The molecular fingerprint of mice with prion disease is different from that of healthy mice. Since different diseases perturb different combinations of networks, these fingerprints may serve as organ-specific markers of disease, quickly and inexpensively assayed in less than a drop of blood. “We’ve used deep transcriptome analysis for about 50 different human organs and 50 different mouse organs,” says Hood. “We have organ-specific proteins for all of those. We’ve termed it the organ-specific blood biomarker strategy.”
Hood nixes the idea of a virtual patient, at least in the foreseeable future, because so much of cell biology is still a mystery. But biosimulation will be essential in getting there. “We want to use perturbations of various sorts to test our network models, and as those models get more complicated, it isn’t obvious what perturbations you should use until you can go into the computer and test a gazillion things,” Hood explains. “Simulating complexity is a key point.”
Trey Ideker, PhD, a Hood protégé, brings up an interesting point when he asks, “The cell is a complicated beast. How do we get those pathway models to begin with, given that there’s really not much in the literature about how the cell works? How does one systematically map those pathways to the point where you can do all these simulations?”
Now associate professor in the departments of Medicine and Bioengineering at the University of California–San Diego, Ideker also heads the systems biology lab. He became part of history when he went to work in Hood’s lab at the University of Washington and later coauthored a Science article with him on how to do systems biology. The dearth of data is one reason the Ideker lab emphasizes basic cell biology research, and why he thinks it makes more sense to do research than to curate the literature. “The amount of data that will come out this year about my favorite pathway, DNA damage response, is greater than all the data that has ever been collected in all of human history about that pathway,” says Ideker. “Yeah, curate the literature. Definitely read the literature and try to incorporate it. I’m just saying what we don’t know is a lot greater than what we do know, so we need to avail ourselves of the technologies that are going to help us systematically screen the system and build these models for us.”
P4 medicine (predictive, preventive, personalized, participatory) will “utterly transform the business plan of every single sector of the healthcare community.”— Leroy Hood, MD, PhD, Institute for Systems Biology
Funding for the Ideker Lab includes grants from Unilever, Pfizer, The David & Lucile Packard Foundation, National Institute of Environmental Health Sciences, National Institute of General Medical Sciences, National Institute of Mental Health, National Science Foundation, and Microsoft.
The 20 members of the Ideker Lab are working on network-based biomarkers, models of signaling and regulatory pathways, using network models to diagnose disease and map protein networks in the pathogenesis of diseases such as HIV, neuroAIDS, malaria, and herpes. Ideker mentions an upcoming paper on a network-based approach developed by Thomas Kipps, MD, of the UCSD Moores Cancer Center for diagnosing chronic lymphocytic leukemia. Then he weighs in on the Genome-Wide Association Studies (GWAS) controversy, about which he coauthored a paper in PLoS Genetics. For Ideker, the GWAS concept illustrates the futility of trying to explain disease in terms of single nucleotide polymorphisms. They’re trying to find new gene associations without taking into account how these genes are actually wired together, says Ideker, “so it’s not at all surprising that looking at just one gene at a time isn’t going to tell you much.” What we feel is going to work a lot better, he says, is to look at the blueprint to see if your genes A, B, and C are in the same part of the network. You really need to be informed by pathways and circuits in the cell, and then you can ask pathway-level questions and make pathway-level inferences.”
Systems biology hasn’t made it on Oprah yet, and more people are probably still working on individual genes and proteins than on disease networks. But there is little doubt that the game has changed. The avalanche of new data is unprecedented, and it’s costing less to generate it, as are the abilities to search and sort it, build virtual models, biosimulate, and predict it.
“Now that we recognize we have to have a systems approach, how do we do that as efficiently as possible?” Bangs asks. “These kinds of systems will take us all the way from discovery through clinical care. I think systems biology will eventually have a role in clinical decision systems, patient decisions, and patient education.”
Don’t bother asking Hood whether systems biology will have the biggest impact in therapeutics or diagnostics, because he’ll tell you that systems approaches are going to “transform everything.” He envisions revolutionary change — connecting systems biology with P4 medicine (predictive, preventive, personalized, and participatory), a notional successor to personalized medicine. Hood predicts that systems biology plus emerging measurement technologies and new computational and IT tools are going to catalyze P4 medicine. “P4 medicine will, in the next 10 years, utterly transform the business plan of every single sector of the healthcare community. It’s going to lead to a digitalization of medicine and an ability to deduce disease-relevant information from a single molecule or single cell. I forecast a time when escalating costs of healthcare are going to turn around sharply and go down rapidly to the point we’ll be able to export P4 medicine to the developing world.”
G. Steven Burrill is chief executive officer of Burrill & Co., a life sciences merchant bank in San Francisco focused on companies involved in biotechnology, pharmaceuticals, diagnostics, devices, and human healthcare. His take on systems biology is even more optimistic.
Healthcare delivery today is not much different than it was 2,000 years ago, Burrill comments. “We largely wait for people to get sick, we try to do something, and they live or die.” Envision a world in 2020, he says, in which you get up in the morning and you spit on a microfluidic sensor on your Black-Berry or your iPhone, which telecommunicates up to the magic computer in the sky, much like a GPS system. “Healthcare will change dramatically by integrating the technology around us in a way that allows us to be in the semi-real time business of understanding what we’re calling personalized medicine, and from that we will get to prediction.”
That won’t happen overnight.
“This medicine we’re coming to understand, which some people call personalized, and I call P4, is in for very tough going, because the current medical structure’s going to be very, very conservative,” Hood says. “I think we’re going to have to take entirely new approaches to pioneer the transformation by working with medical centers on demonstration projects.”
Should payers and purchasers be excited about systems biology?
“They probably don’t know it yet, but they should be,” Bangs quips. They’re already doing what they call predictive modeling to identify who is most likely to end up in the emergency room next week, but they may not know exactly how to help that person other than give him or her more attention. If we can simulate a population of people to give us a good idea of their future health risk, he says, and what might best be done for them, then intervention can start to become much more personalized.
Sounds like a business opportunity, and Bangs sees some sort of collaboration with payers and healthcare purchasers within the next few years. “It’s more a question of how and where to best apply it.” In the end, Burrill thinks, what’s mind-blowing today will be commonplace tomorrow. “We’ll be doing systems biology for the next hundred years, so it’s not like we’re going to get it done and move on. It’s understanding how life works.”
Marc Kirschner, PhD, can tell you how life works, especially when it comes to the cell cycle and signaling in cells, his main research areas. He heads the systems biology department at Harvard Medical School and takes the long view of this new approach.
“In some ways, systems biology is like molecular biology, which used to be in departments that did that sort of thing,” Kirschner says. “But molecular biology was so successful that in a way it’s almost disappeared, because it has become part of biology. Maybe 50 years from now there won’t be any departments of systems biology, they’ll just be incorporated into the rest of science.”