The Alpha Project

One day, genomic data will be translated into language that can be used to find new diagnostic and therapeutic targets for disease. Computers will mine DNA codes to build nanomachines, and "smart fabrics" will contain sensing capabilities modeled on living things. So says Shankar Shastry, chairman of electrical engineering and computer sciences at the University of California at Berkeley. "Bio is my bet on where the new set of glamour technologies will be," he predicts. But even the small steps that begin such giant leaps are hugely difficult to take. Witness the Alpha Project. Courtesy of The Molecular Science Institute Roger Brent It's the brainchild of Roger Brent, president and research director of the Molecular Sciences Institute (MSI) in Berkeley, which was founded seven years ago by 2002 Nobelist Sydney Brenner as an independent, nonprofit research organization. Brent, a widely published former genetics professor at Massachusetts General Hospital and Harvard Medical School, joined MSI in 1998. Years before that, he began lobbying for what has become the group's signature effort, the Alpha Project. Its goal is to gain the ability to predict the behavior of what he considers to be the best understood signaling pathway in biology: pheromone response (PR) in the brewer's yeast, Saccharomyces cerevisiae.1 Essential to reproduction in this single-celled organism, PR is prototypical of regulatory networks in higher eukaryotes, including humans. But to fully characterize the flow of information coming from without and within the cell requires quantification. In turn, this demands an analytical framework that tells the researchers what to measure and how to make those measurements. The Alpha Project's funding, $15 million (US) over five years, began arriving last summer from the National Human Genome Research Institute. MSI is the core facility, with participation from scientists at four major US universities. The coalition will claim victory when the pathway can be modeled, perturbed, and outcomes can be accurately predicted. During the Alpha Project's first annual research symposium in December at UC, Berkeley, Brent said: "We're going to say that the ability to predict is tantamount to understanding." During a later interview in MSI's crisply geometric lab in downtown Berkeley, Brent noted that in systems designed by humans, the engineers know the components and their relationships. That's what most differentiates human design from nature's way of entities "blundering around in the dark," as he wryly put it. A key aspect of such "blundering" is variability in behavior, a challenge that requires information extraction from individual cells. This variability highlights the twin objectives, analysis and measurement, that underpin the Alpha Project's overall goal of predicting PR. DETERMINING VOLUMES AND RATES The obvious analytical framework is to chronicle reactions that entities undergo, but such a verbal narrative differs from the mathematics in significant ways, says Brent. One method of dealing with that problem involves physically tracking the estimated 50 million protein molecules in a yeast cell as they diffuse in space and bump into each other. Whenever a reaction is possible between two molecules that meet, statistical tables could be used to find the probability of phosphorylation. This approach is conceivable with supercomputers, but a drawback is that most of the reactions thus identified would be inconsequential. Another avenue is to use qualitative information for predictions, but that undermines the Alpha Project's mandate to gather quantitative data. Knowing volumes and rates will provide the ability to observe new phenotypes, leading to identification of mutants that, in turn, will identify relevant genes. In short, better quantification should lead to better qualitative data. Making the measurements is the major rate-limiting step. Protein phosphorylation can be detected by mass spectrometry as a change in mass but the question is, can volumes and rates be determined? If so, the spectrometer could become the equivalent of the voltage meter on an electrical current. The first challenge will be to make a census of phosphorylation events that connects each specific charge-to-mass ratio to a peptide encoded by a particular gene. Next, the rate of its appearance must be measured in response to a stimulus. Variability in the behavior of genetically identical cells must be addressed by fluorescently tagging individuals. Courtesy of The Molecular Science Institute Alejandro Colman-Lerner A NEW LEXICON Given these issues, it's not surprising that simulation of the network is still in its early stages. "Right now, we can't even pretend to prediction at all," acknowledges Larry Lok, one of MSI's biological modelers. He and Alpha Project coinvestigator Alejandro Colman-Lerner are working on an initial step in the PR pathway, which is dissociation of a G-protein complex at the membrane receptor site. Engineer and biologist are devising a mode of communication that can put biological ideas into words and figures that the computer can use to simulate such reactions. The two men engage in intensive interchanges concerning what should happen in the simulation and which of the millions of data bits should be "dumped" for viewing. Colman-Lerner has to determine the reactions, such as binding and inhibition, but Lok can help to figure out unexpected dynamical ramifications. They often sit together at a blackboard, drawing cartoons to explain what needs to be included in a given part of the simulation and what effect each inclusion will have on the whole scenario. Numerous trips are made back to the blackboard for further refinements. Courtesy of The Molecular Science Institute Larry Lok In a typical simulation, a species produces features that are sent along the network. Reactions that permit dimerization are recognized, the resultant union creates new species with new features to send along the network, and so on. A main assumption is that dimerization is based on mass, which means that the "on-rate" declines with size but decomposition or the "off-rate" remains unchanged by mass. Lok says one of his next challenges is to find some way to simulate steric constraints, or the effect on reactions of spatial relationships between atoms in molecules. The problem is that these geometries are largely unknown. Lok demonstrates a basic simulation, beginning with 3000 "A" and 1000 "B" entities or mols, each with binding sites and a prescribed weight. The computer runs the reactions in one second, and a viewing program shows the numbers of As and Bs declining in direct proportion to the creation of ABs. All the reactions are dumped out in XL, a computer language so verbose and cumbersome that the need for a script usable by humans rather than machines becomes immediately clear. A real simulation is far more complicated than Lok's demo. For example, the PR pathway is both positively and negatively regulated. It tries to turn itself off, while even more actively trying to turn itself on. "The pathway itself regulates the negative regulation by inducing it, strangely enough," says Colman-Lerner. Further complicating the pathway is cell cycle control on PR. "So, there are multiple arrows going in both directions, trying to keep the pathway at some level of activity. This is what we want to measure." Through the conversation of pheromone signaling, the two sexes of yeast cells, "a" and "alpha," fuse into a diploid, explains Ira Herskowitz, a UC, San Francisco, biochemistry and biophysics professor. "You could also think of the pheromone response pathway as a big macromolecular complex, and there's a lot of coming and going.... What's happening is building structures and taking them apart." While Herskowitz admits to a scant background in quantitation, he supports attempts to model the PR pathway and offers a reminder that summarizes a major challenge for the Alpha Project. "The pheromone response pathway was one that evolved," he says. "It wasn't designed." Steve Bunk (sbunk@the-scientist.com) is a contributing editor. 1. H.G. Dohlman, J.W. Thorner, "Regulation of G protein-initiated signal transduction in yeast: Paradigms and principles," Ann Rev Biochem, 70:703-54, 2001. Courtesy of Bill Sherwood AN ENGINEER GOES BACK TO SCHOOL What could prompt the Massachusetts Institute of Technology's head of mechanical engineering to take freshman biology? For Nam Suh, it was a theory he'd developed 20 years ago to create more efficient machines, software, and manufacturing processes that could be applied to understanding life. But he needed a solid biology foundation. In the 1980s Suh challenged the notion that engineers could learn design only by experience and developed a systematic method for planning man-made systems called axiomatic design. This top-down approach relegates a system's functional requirements down to its physical components. Similarly, the things a living organism needs to do (eat, move, reproduce) can be mapped to hierarchies of constituent parts in the physical world--organs, cells, proteins, genes. When applying the principles of axiomatic design, says Suh, well-designed systems take on a periodic complexity. This means that a period of rest or resetting must interrupt function. From sleep to mitosis, migration to membrane repolarization, periodicity in biological systems can be introduced by time, geometry, thermal cycles, electrical charge, or simply process completion, as in checkpoint control. If periodicity stops, the system switches over to what Suh calls combinatorial complexity, and breakdown follows. People experience this all the time: "If your PC goes down you have to reboot it," explains Suh. "That's because if you keep running a computer for a long time, pretty soon combinatorial complexity takes over." Ravi Iyengar, pharmacology chairman at Mount Sinai School of Medicine in New York City, penned a paper four years ago on biological complexity1 that piqued Suh's interest. He now collaborates with Suh and Columbia University's Michael Sheetz to apply axiomatic design to cell motility. Iyengar says he forced Suh to take freshman biology to get them on the same page. Courtesy of Nam Suh Nam Suh Iyengar says he's uncertain how well periodic complexity translates to living systems. "In this sort of biology there may be parallels in a macro term, but really cellular complexity is more of a micro problem or something perhaps in the middle." Suh says that many biologists trying to gain a system-wide understanding are wasting their time because they are looking at complexity the wrong way. "Most biologists are so busy thinking about molecular interactions and molecular behavior that they don't really think about system issues." These interactions may not be contributing relevantly to the functional requirements of the system. Iyengar agrees. "I'm very convinced that his one idea of sorting through real complexity versus apparent complexity is very important." -- Brendan A. Maher 1. G. Weng et al., "Complexity in biological signaling systems," Science, 284:92-6, 1999.

The Alpha Project

Interested in reading more?

Become a Member of