|
THINKING LIKE AN ENGINEER
Although Arkin is certainly not the only biologist trying to understand the cell as a system—as opposed to looking at it gene by gene or protein by protein—he is tackling the problem with a broad range of tools and techniques. Mathematical modeling, network analysis, computer simulation, laboratory experiments—he and the 20-plus postdocs and students in his laboratory are using anything that seems to work.
How else, he asks, can you see the underlying phenomena? Consider the neutrophil, a type of white blood cell that's critical to the body's defense against disease. This organism lacks eyes, hands and a brain, yet through chemotaxis—the ability to follow subtle chemical traces in the environment—a neutrophil can lock on to an invading pathogenic bacterium, track it down as it darts among the surrounding red blood cells and destroy it with the accuracy of a heat-seeking missile. "This is an incredible navigation system," says Arkin. "But the number of proteins involved in that system is huge—in the hundreds, at least. We're not going to understand that cell until we get some way of ordering this information. We need a theory to encompass it all. We need the data structures to hold it. And we need the knowledge representation that allows us to query it in clever ways and help us learn how it fits together."
One way to pursue this goal, he says, is to use a wide variety of organizing principles. He embraces fields such as electrical engineering and communications theory, in which practitioners have long since evolved ways of analyzing complex networks.
Arkin holds up a printout: "Here is a network diagram for sporulation in Bacillus subtilis"—a harmless bacterium that, like its deadly cousin B. anthracis, the cause of anthrax, will sometimes turn itself into a hardy spore under conditions of stress. How does it make that decision? "Look at this, this and this," he says, pointing to clusters of interactions that seem to have a similar structure. These clusters are examples of what he and his group have dubbed a regulatory "motif." They all have a promoter to guide the expression of two genes: one is the activator for a process and the other is the inhibitor. If you think like an electrical engineer about such a system, he says, you can analyze how its overall behavior will arise from the push and pull of those two genes.
"It turns out that this regulatory motif has the ability to be a switch, a pulse generator or an oscillator," he says, depending on how the push and pull are balanced. Furthermore, he adds, "You get the same pattern of regulation from evolutionarily unrelated proteins. So it's the structure of the network that's important, not the identity of the proteins."
With that insight, you can model these subnetworks as components in an electrical circuit, or perhaps as active nodes in a communications web. "When you put those oscillators and switches back into the network," Arkin says, "you can determine what its possible dynamics are at a higher level of abstraction. So if I were to perturb the cell by a change in the outside environment, I can begin to analyze how the signal is processed through the network" to trigger sporulation or any other change in behavior. Though the same results could be derived from a full physical model, he adds, this high level of abstraction, if done correctly, could give us a much clearer picture of how cell behavior arises.
Of course, admits Arkin, this high-level, functional approach to cell behavior has plenty of skeptics, who wonder if electrical engineering bears anything more than a vague resemblance to biology. Nonetheless, he says, he and his students have already used the technique to produce some interesting preliminary models of B. subtilis and several other organisms. In addition, he and his group are collaborating with many other laboratories on a related approach to organizing knowledge of the cell: a kind of online biology network library and cell-simulation tool known as BioSPICE.
Arkin doesn't claim to be the originator of BioSPICE, just one of its earliest and most vocal proponents—although he does take credit for the name. That came from an electrical-engineering tool called SPICE, the Simulation Program for Integrated Circuit Evaluation, which does pretty much what its name suggests. He says, "when I first started pushing the idea, there were already a couple of simple programs out there for cell simulation. But I meant BioSPICE to be more than that." Indeed, he and his allies argued that the system should integrate every form of molecular data available, from DNA sequences to protein structures and beyond—and then provide a seamless interface for any cell simulation or network analysis a researcher wants to write.
That's obviously a huge undertaking, says Arkin, which is why "it's now DARPA BioSPICE, not just my BioSPICE." In September 2001, when the Defense Advanced Research Projects Agency started funding multiple laboratories to work on various pieces of the project, Arkin became one of 19 or so principal investigators. Nonetheless, he says, BioSPICE remains "a major core" of his lab in Berkeley.
He calls up a portion of the program on his laptop. "At the heart of BioSPICE is this pathway diagramming tool, which is our interface with working biologists." He shows the "cartooning" view, in which a stretch of DNA is a simple straight line, the various genes of interest are yellow ovals embedded in the line, proteins are free-floating blobs and various processes among these components are specialized lines. "But even in a cartoon there are issues of what you want to include," says Arkin. "For example, the central dogma of biology is that DNA gets transcribed into messenger RNA (mRNA), which gets translated into a protein. But if you're not going to be talking about the RNA in this experiment, you don't want to show it."
Indeed, in this particular view, the complex process of gene expression has been collapsed into a simple arrow connecting each gene to the appropriate protein. "But the software has to know that there is still an RNA in there someplace," says Arkin. If need be, in fact, the software allows the user to drill down and show not just mRNA, but its binding sites, cleavage sites, terminators and all the rest. It can go even further than that. "For example," says Arkin, pulling up another display, "here's where someone has drawn two proteins binding to DNA. But what does that mean? It could be that this guy binds first, then that guy. It could be the other way around. It could be that that guy binds, and he prevents this guy from binding. And so on. We have to allow users to put that kind of information down—even if it's just to say that we don't know how these proteins interact.
"So that's one of the issues we're really struggling with," Arkin says: "How do we represent all that information? And how do we do it so that biologists who don't know anything about computer models can navigate around this information with ease?"
A SOCIAL EXPERIMENT
Along the way, Arkin and his team have also been struggling with an unusually complex social dynamic. Between the programmers and the various flavors of scientists and engineers, his group spans 10 different disciplines: chemistry, chemical engineering, molecular cell biology, mathematics, physics, computer science, electrical engineering, bioengineering, mechanical engineering and bioinformatics. His lab has expanded from fewer than half a dozen members to more than 20 within the past two years.
Everyone is still working through the culture clash. "You can imagine that a student coming in gets very disconcerted," says Arkin, "because the guy next to her is from a different field and looks like he knows infinitely more than she does. But he's thinking the same thing about her. So they both look at each other and say, 'My God, I'm unqualified for this job,' and there's that uncomfortable period of learning."
It probably doesn't help that members' offices are scattered over all four floors of Berkeley's Calvin building while their new lab space is being constructed. Nonetheless, they generally give Arkin high marks for his ability to build team spirit. He does it partly through his weekly group meetings and discussion sections, they say, but mostly through his rapid-fire personality.
"Adam just creates an intellectually stimulating environment," says postdoc Christopher Rao, who is doing a comparative analysis of how chemotaxis is controlled in Escherichia coli and B. subtilis. "It's not especially well-oiled, but this is exactly the kind of multidisciplinary research group that everybody says we need in biology."
Arkin's background is uncommonly eclectic. During his high school years, for example, the one-time trash collector became interested in the brain. "It seemed like a natural progression," he explains, "the brain being an organic kind of machine. I found my way into a programming job in the neurosurgery department of a hospital in New York, where I learned my first laboratory skills. I had told them I knew all these computer languages—though I didn't. I had to learn them, fast!"
Then, as an undergraduate at Carleton College in Minnesota, Arkin switched to physical chemistry—a field that he pursued through his early graduate-student years at the Massachusetts Institute of Technology (MIT)—until he found himself working with a biologist on a study of one of the proteins involved in photosynthesis. That was interesting enough, he says, "but when I began to look at the endogenous pathways to control this entire process, I was just captured. And I started thinking, 'That's how it actually works! And I'm studying only one protein?' " He wanted to understand it all, he says. In those days—the early 1990s—biologists were finally beginning to accumulate enough data to make that fantasy seem possible.
"Being a geeky kind of guy," says Arkin, "my hypothesis was that there had to be some form of control and computation inside a cell." So in 1992, Ph.D. in hand, he left MIT for a postdoc with Stanford University chemist John Ross, who was studying how to make chemical reactions carry out computer-like operations. Just as he was finishing up his postdoc in 1995, he noticed a new paper about lambda phage (a virus that infects E. coli) by Stanford's Harley McAdams and Lucy Shapiro. "It was actually a very, very nice paper," he says, "but, young upstart that I was, I found lots of faults with it—most of which were beside the point of the paper. So I called them up and started an argument with Harley. It was a great argument! We had so much fun that we decided I should migrate to do a postdoc in their lab."
"Harley was an amazing mentor," says Arkin, "and so was Lucy. Harley was working very, very hard to pin down everything he possibly could about lambda phage. Together we came up with this theory that genes have to be expressed stochastically (that is, with lots of statistical fluctuation in the rate of expression), given the low number of molecules that control them."
That finding was an eye-opener for many biologists, and it may well have been what got Arkin his next job at the Lawrence Berkeley Laboratory, located on the hillside just above the University of California campus. He arrived there in January 1998, joined the university's faculty in July 1999 and became an HHMI investigator in October 2000. Perhaps more important, however, the finding crystallized his current commitment to systems biology: "What came of working with lambda phage," he says, "was a renewed understanding that everything was connected—that there is order and principles to cellular control."
His hope, Arkin says, is that this kind of systems work will eventually transform molecular biology into a kind of cellular engineering—a discipline in which practitioners can predict, control and design cellular materials as confidently as traditional engineers create, say, a new aircraft.
"Just look at the following Holy Grail problem," he says. "Given a known genetic predisposition of human beings, and the network that controls a cell, predict the best place for a drug, or a combination of drugs, to move that cell from sickness to health. You want to be able to use cellular engineering to cut down the time to find drug combinations—or perhaps even to tailor the drugs to individuals.
"There are other practical applications," he adds, "for example, designing an organism that will metabolize, say, a dangerous heavy metal like mercury into a compound that we can immobilize, extract and put someplace safe."
"It's an interesting quest, in a knightly sort of way," says Arkin, "and I think I want to take that on."
Download this story in Acrobat PDF format.
(requires Acrobat Reader)
Reprinted from the HHMI Bulletin,
September 2002, pages 20-23.
©2002 Howard Hughes Medical Institute
|
|
|
