The receiving line includes distinguished professors and humble graduate students, the well known and the unknown, scientists who spend their days peering at computer screens and others who spend theirs peering through microscopes, representatives of government agencies, a sampling of corporate sponsors, and a contingent from a respected scientific publisher. It’s the 2002 opening reception for the second annual meeting of the Alliance for Cellular Signaling, a consortium of investigators with a common interest in the language of life, and the man everyone is waiting to greet is Alfred G. Gilman, the Alliance’s creator, director, taskmaster, and muse.

Regal in his formal dark suit and crisp white shirt, Gilman looks the part of the elder statesman as he greets his guests. Launched in September of 2000—two decades after he added “G_s” to the lexicon of signaling molecules—the “AfCS” is Gilman’s idea for making sense of the multivolume tome the dictionary has become. Over a 10-year period, the 80-odd scientists who currently make up the AfCS intend to compile a comprehensive “parts list” for every signaling pathway in two types of mammalian cells: the mouse B lymphocyte and

the myocyte, or muscle cell, of the mouse heart. But that’s only the beginning. The AfCS also hopes to understand how these proteins interact to process information, “the big question of how they all work together as a system.” Then, armed with this information, they plan to design a computer model of each cell that will re-create the network of interconnected signaling pathways and use this “virtual cell” to explore the dynamics of the signaling circuitry as well as aid in the design of new drugs to correct defects in that circuitry. Oh, and along the way, they hope to change fundamentally the way science is done in the twenty-first century.

“Sequencing the genome is enabling us to think about bigger questions in unbiased, nonhypothetical ways,” Gilman noted in an interview at the time of the launch. But solutions to bigger questions, he believes, are more likely to come from the collaborative efforts of many scientists, pooling resources, sharing data, and developing the analytic tools to manage and mine those data than from the time-honored style of research based on individual scientists working by themselves, for themselves—solutions, Gilman argues, that can be facilitated by modern communications technologies like the Internet.

To build a virtual cell, therefore, he has built a virtual laboratory. Eight research teams, located in San Francisco, San Diego, Pasadena, and Dallas, carry out the actual experimental work and develop the software needed to store and analyze the data they generate. An additional 41 participating investigators serve on committees that oversee the work of the Alliance laboratories. Everyone communicates regularly via teleconferencing and the Internet, enabling them to coordinate their efforts as if they worked in the same building. Finally, approximately 1,500 subscribing members will contribute by authoring so-called Molecule Pages, peer-reviewed descriptions of information about signaling molecules the AfCS will catalog and publish as an electronic database freely available to all members of the scientific community.

Big science takes big money. The annual budget for the project totals around $10 million, much of it coming from a $25 million “glue grant” awarded to the AfCS by the National Institute of General Medical Sciences, as part of an initiative to support and encourage such collaborative endeavors. The balance will be made up by funds from the National Institute of Allergy and Infectious Diseases, the National Cancer Institute, paid subscriptions from a consortium of pharmaceutical companies, and several private donors. In addition, a group of biotechnology companies contribute to AfCS research by supplying equipment and reagents essential to specialized, state-of-the-art techniques.

“This is not an easy operation to run. You could think of it as a small company spread all over the place,” Gilman concludes. “Participants do not have proper, conventional academic motivation. We’re not providing money for their laboratories or publication opportunities. People are part of this because it’s an interesting—and important—project.”

The 2001 meeting was all about organization—the design of experiments, the standardization of procedures for recording, transmitting, and analyzing data. At this meeting, AfCS investigators will unveil the first actual results of their efforts. The mood is expectant, excited. Members greet one another as effusively as if they were attending a high school reunion rather than a scientific meeting. It’s been a hard year, everyone agrees. And the biggest problem hasn’t been the experiments or the data analysis, but the struggle to get people more accustomed to competing with one another to collaborate instead.

“It’s been a process of organization, as well as science, sorting out the hierarchy,” says one member. Another, one of the computer experts responsible for organizing the Alliance database, agrees, noting that despite their reputation for fussiness, “bench” scientists are anything but consistent when it comes to record keeping. “One of the

hardest parts has been teaching scientists to think about their data before they do an experiment, rather than afterward,” he explains. “They’re not used to standardized procedures. You have a notebook and one guy wants to keep it his way and another wants to use his own way. One wants to write in red ink and one in blue ink”—or the electronic equivalent anyway. A third complains that some tasks still haven’t been converted to an electronic format and must be carried out manually. “It’s like a big ocean liner, out in the middle of the ocean. At one end there’s an engine powering the boat along, and at the other there’s still some guys rowing,” he says as he spears another shrimp roll from a passing tray. “And I’m just glad it floats.”

Gilman sums it up: “We are both experimentalists and experimental subjects. This is an experiment in how to do science as well as an opportunity to find out about signaling systems. There are people watching to see if we can make this model work. The world, in general, is a little bit skeptical of big science. And the world is watching. We want to deliver what we promise.”

In 1992 the journal Trends in Biochemical Sciences published an update to the “black box” insulin cartoon that diabetes expert Ron Kahn found so compelling. The new version also depicted a cell and its insulin receptor. But now instead of black shading, the area inside the cell contained a bewildering array of insulin substrate proteins, second messengers, kinases, adaptors, glucose transporters, and transcription factors, connected by a labyrinth of intersecting lines and arrows. At the bottom the new caption reads, “Now we really understand insulin signaling.”

Do we?

Over the past four decades, biologists have compiled an impressive list of signaling molecules, from the chemotaxis proteins of E. coli to the plethora of insulin-related proteins and the dozens of neurotransmitters ferrying messages in the brain. They have demon-

strated how the loops and notches of protein domains can link protein to protein or pathway to pathway. They have cataloged patterns and outlined the rules of cellular grammar. They have demonstrated how signaling mechanisms can connect environment and genome, strike a balance between excess and shortfall, reconstruct as well as regulate.

The achievements of signaling research are representative of our success at analysis—tearing things apart—and our urgent need for synthesis, as scientists have also come to recognize that cell-cell communication is more than linear information transfer. In the living organism, “signaling pathways interact with one another and the final biological response is shaped by the interaction between pathways.” Like the fiber optic systems that route telephone calls or the hardware that shuttles data inside your computer, biological communication is the product of integrated circuits operating as a network, and the collective behavior of that network, not the sum total of the activity in myriad individual units, is what determines the cell’s responses to external signals.

“The term ‘pathway’ implies a linear structure—you flip a switch and a particular biological process happens at the other end. And that’s clearly flawed,” says cancer researcher Gerard Evan. “Signaling pathways can never be linear because they would never work,” he continues. “There aren’t enough genes to encode all the combinations you would need. Any time you couple two pathways and it confers a selective advantage, that connection will be retained. So what you get after billions of years of evolution are connections.

“We’re primates, so we like to put things in boxes. But evolution is like going into your basement and sorting through the rubbish to see if anything is useful.”

Evan cites the transcription factor Myc as an example of a protein that has been favored by evolution because it could be recycled, mixed and matched with other proteins to form multiple signaling pathways. “Myc is an integrator, bringing together all the things you

need to proliferate. Any gene coupled to myc that enabled an organism to build tissues more effectively would be selected for. So you get a bewildering proliferation of pathways.

“Process comes from the way pathways interact, not specific proteins,” Evan concludes. But it’s one thing to acknowledge the importance of interactions and networks, and another thing entirely to investigate signaling from a network perspective.

“Complexities can only be understood by constructing quantitative, often mathematical models,” writes former Nature editor John Maddox in his book What Remains to Be Discovered. According to systems biologist Eric Werner, scientists realize three benefits when they construct such models to explain complex systems. First, models allow researchers to evaluate their progress, because the development of models exposes gaps in their understanding of biological systems. Second, models can be used to inform future experimentation: “Models … can be used to plan experimental strategies,” while “cycles of modeling and experimental validation gradually result in the convergence of the model’s predictions with the measured parameters of the natural biological system.” Finally, Werner argues, the quantitative rigor demanded by model building exposes fuzzy thinking; as he puts it, models “force a new perspective on the subject matter. One can no longer tolerate intuitive, vague models. One is forced to look at the consequences of theoretical assumptions.”

Unfortunately, as Maddox notes, biologists have been slower than other scientists to appreciate the value of model building. Their reluctance stems in part from the fact that many are unfamiliar with the specialized mathematical and computer skills that modeling demands or are disconcerted by the idea of experiments conducted “in silico,” on computers instead of at the lab bench. But the cult of the individual researcher is also a barrier. An integrated biology that seeks to combines database management, data mining, and network analysis with experimentation, especially experimentation that involves state-of-the-art molecular biology and imaging techniques, is “big

science,” made possible only by pooling time, effort, and expertise. If biologists are to construct models as well as perform experiments, it will require not only a paradigm shift, “refocusing on systems, rather than components,” but also a cultural revolution—an overhaul of the field that Gilman and his collaborators hope to pioneer.

Spoken language is more than strings of sounds or even combinations of words. As architect Christopher Alexander says, “Each word carries the whisper of the meanings of the words it is connected to”; that is, the meaning of the entire sentence depends not only on the words themselves but also on the relationship between words. In the living language of life as well, the relationships between chemical words add a subtlety that transcends the individual elements. We can parse pathways into proteins, but to understand the coordinated behavior of real organisms, we must also understand how meaning emerges from the confluence of proteins and the interactions between pathways.

If you’re wondering what all this number crunching and paradigm shifting has to do with you, there’s a practical benefit to what the AfCS is attempting to do as well—it will be good for your health. Chemotherapy, Paul Ehrlich’s idea that cures could be found in chemical agents with a specific affinity for the “side chains,” has been one of the most productive concepts in medicine, and the interactions of experimental compounds and signaling proteins, particularly receptors, represent the cornerstone of modern drug discovery. Today, the overwhelming majority of drugs in the modern pharmacopoeia act by talking to cells in their own language. The discovery of new signals and new receptors, as well as the intracellular relays they engage, has offered new targets for pharmaceutical researchers, opportunities that are already beginning to yield more effective and less toxic medications—and, at the same time, have revealed the pitfalls of concentrating on individual receptors or kinases instead of the larger picture, particularly the interrelationships between signaling pathways.

The general public may view cancer as “a modern-day plague” and oncologists may see it as “a legion of recondite diseases whose diversity complicates therapy,” but to a cancer patient, it is “a terrifying alien entity invading his body and treatable only with medicines of medieval harshness and dubious efficacy,” writes Gerard Evan. Indeed, for decades, cancer treatment has focused on eradicating tumor cells by brute force, with surgery, radiation, or powerful drugs; “slash, burn, and poison,” breast cancer authority Susan Love calls it. The side effects of bludgeoning cancer into submission can leave patients wondering if the cure isn’t as bad as the disease.

But in 1998, Genentech introduced a different kind of cancer drug, an agent that attacked cancer by silencing its propaganda machine rather than by mounting a brute force assault. The drug, Herceptin, is a monoclonal antibody to a domain in the protein encoded by the her2/neu gene, related to the receptor for epidermal growth factor. Malignant cells in some breast cancer patients contain extra copies of the her2/neu gene; these cells proliferate extravagantly because they’re saturated with extra receptors as a consequence (as many as 1.5 million receptors per cell, compared to about 50,000 on normal cells), leading them to imagine they are being bombarded with growth factor. By binding and blocking these receptors, Herceptin turns down the volume—without harming normal cells.

In 2001, Novartis followed this success with Gleevec, another drug that works by talking cancer into remission. In patients with chronic myelogenous leukemia (CML), Gleevec silences an aberrant kinase, Bcr-Abl, created when segments of chromosomes 9 and 22 accidentally switch places. In patients with a rare type of intestinal cancer, gastrointestinal stromal tumor, Gleevec blocks a mutant receptor tyrosine kinase called c-kit. Finally, it also inhibits the platelet-derived growth factor receptor tyrosine kinase; as a result, clinical trials are evaluating its effectiveness in cancers with mutations in this receptor. Gleevec is breathtakingly effective—90 percent of patients

with CML who take the drug in the early, chronic stage of the disease have their cancers go into remission, as do 60 percent of patients with gastrointestinal stromal tumor. In addition, because it targets defects specific to cancer cells, Gleevec, like Herceptin, spares normal cells, minimizing the incidence of side effects.

A third signaling oriented drug, AstraZeneca’s Iressa, was approved in 2003 for the treatment of non-small-cell lung cancer. Like Herceptin, Iressa shuts off the relentless screaming of a corrupted epidermal growth factor signaling pathway, in this case due to mutations in the growth factor receptor itself. Rather than smothering superfluous receptors, however, this agent works inside the cell to prevent the EGF receptor from phosphorylating itself. Without the addition of phosphate, the receptor can’t communicate with its intracellular signaling relay, and growth factor–induced responses grind to a halt.

“The war on cancer was started by Nixon, but we’re only making progress now,” says cancer researcher Philip Beachy. “As we understand more about the cells involved in tumors we can target those cells instead of using blunt instruments that kill all dividing cells. In the future, oncology will have a different set—a more specific set—of therapeutic tools. Instead of cytotoxic agents, we’ll have mechanism-based agents.”

Continued progress in developing a “more specific set of therapeutic tools” depends critically on a better understanding of the interlocking signaling mechanisms governing the progression to malignancy. A challenge, yes, but not an insurmountable one, argues another cancer expert, Gerard Evan. “We have only two options to collapse the cancer platform,” Evan writes. “One is to attack the lesions that drive tumor cell proliferation”—such as growth factor receptor mutations. “Alternatively,” he continues, “we could reinstate the defective apoptosis, whereupon the tumor cell should die from the apoptotic depredations inflicted on it by its driving oncogenic lesions.” He acknowledges that this second option is currently

hampered by our relative ignorance of the genetic accidents that derail the suicide program in cancer cells. Still, he points out, we may have already begun to correct the failure to die in spite of ourselves, noting that drugs designed to interfere with growth factor signaling not only block the compulsion to proliferate but also eliminate essential survival signals. With a better understanding of suicide signaling—and dozens of antiapoptotic drug candidates under investigation—the future may lead to more deliberate approaches to rebalance cell proliferation and cell death.

Signaling drugs like Herceptin, Gleevec, and Iressa are not the latest miracle cures. Only 25 to 30 percent of breast cancer patients have the defect leading to runaway expression of the HER2 protein that is the target of Herceptin (and not all of these patients respond to the drug, possibly because HER2 signaling is more important in some cancers than in others). An even smaller number of lung cancer patients—about 10 percent—have the precise growth factor mutation targeted by Iressa. Gleevec is highly effective in the early chronic phase of CML, but patients with advanced disease in the acute, or “blast-crisis” phase often fail to respond, and some who are respond initially relapse when their cancer cells spawn another mutation that allows them to ignore the drug. But their unique mechanism of action makes these agents the first real advance in cancer chemotherapy in decades of research. Medications that finally hurt cancer more than they hurt the patient, they are leading oncologists to suggest—cautiously—that drugs like Gleevec might turn cancer into a chronic but manageable disease.

At the same time pharmaceutical researchers are making progress in the war on cancer with new drugs that target signaling proteins, they are running into roadblocks in more traditional areas of signaling research. Take drugs that act by stimulating or blocking signaling at G protein–coupled receptors. This category includes many of the industry’s most successful drugs, from antihypertensives to antide-

pressants. Yet progress in identifying new drugs that target the signaling pathways headed by these receptors has slowed to a crawl, in part, some experts believe, because drug discovery research has focused so intensively on identifying compounds that bind more and more specifically to particular receptors, without considering the networks of intracellular signaling proteins that take over once messages cross the plasma membrane. “When success does occur, many come out with very similar molecules targeting the same receptors. Metaphorically speaking, it is as if everyone is looking under the same lamppost to find the key to the biological problems being considered, because they only feel comfortable in that well-lit environment,” says pharmacologist Michel Bouvier, in a recent roundtable discussion published in the journal Nature Reviews Drug Discovery. “We are still at the stage of identifying the various partners that can be engaged by the receptor, and we are only just getting the first glimpses into their potential roles in controlling signaling selectivity and efficacy,” he adds. “Undoubtedly, this will dramatically change the way we approach signaling selectivity, and will hopefully provide new insights into how we can modulate GPCR signaling in a selective manner.”

That’s why pharmaceutical companies are joining government agencies and the private sector in bankrolling the AfCS—they recognize that the more we know about the complexities of signaling, the more successful they will be at identifying and exploiting new therapeutic opportunities.

The mood is still ebullient when Gilman takes the stage the next morning. “You came back!” he announces and everyone laughs. He leads off with a summary of the AfCS strategy and goals: “How does a signaling mechanism work? What are the components of our system? Who interacts with whom? How does information flow

through the system? And finally, can we end up modeling this system and predict the behavior?” Then, to demonstrate what a tall order this actually is, he shows a video clip, a commercial for the computer firm Electronic Data Systems.

“Some people like to climb mountains,” announces a man who appears to be a sky diver, suspended in midair from his parachute. “I like to build airplanes … in the air!” The clip goes on to depict the thrill of victory he and his colleagues experience from bolting an airliner together at 40,000 feet, interspersed with glowing testimonials from enthusiastic passengers. “When I see that look on a little kid’s face, that’s all the faith I need,” the narrator concludes. “Whooohooo!”

“We are building an airplane in the air,” Gilman says as the lights come up. “As you see, it can be done. And the look on that little kid’s face is all that matters.”

Getting back to business, he assures the audience that “progress has been substantial, especially in the last couple of months.” They’ll be hearing details of the methods Alliance laboratories have developed for preparing and propagating each of the two cell types, he notes. And he explains that much of the data they’ll see will come from so-called ligand screens intended to flesh out the “parts list” for the signaling pathways of each cell and “measure the spectrum and pattern of responses.” Basically, the major goal of this part of the work, Gilman observes, “is to collect empirical information for use at later stages.” For example, investigators want to know if all ligands that work through G protein–coupled receptors behave similarly or if they differ and, if so, how. In addition, the ligand screen will attempt to identify modules, groups of signaling molecules that work together as teams in multiple signaling pathways (e.g., MAP kinases). But no peeking, he adds. “I’m not going to take the fun away from the people who have done the work.”

One by one, representatives of the Alliance labs take the stage and review the year’s work, beginning with the laboratories in charge

of the B lymphocyte, which, according to B cell team leader Henry Bourne, of the University of California, San Francisco, is “alive, well, and rarin’ to go!” Their methods of isolating and tending these cells have been perfected to the point that they can routinely prepare nearly 20 billion per week, he notes. Thanks to this productivity, they’ve screened 43 ligands so far, charting alterations in gene expression, the phosphorylation of proteins. And although AfCS members are committed to working with primary cells (that is, cells removed and cultured from animals, as opposed to established cell lines), just to be on the safe side, B lymphocyte investigators have hedged their bets by setting up and screening a B cell line, WEHI-231. “If primary cells can’t be tractable, the hell with them,” Bourne threatens. “We should play to our strengths.” Ideally, he adds, the primary cells will be the “gold standard”—the definition of how signaling transpires in actual living cells—while the WEHI-231 cells will be the “workhorse,” for large-scale studies.

If anything, the B cell screens have been too successful, Bourne says, generating a “tsunami of data.” As a consequence, this team’s greatest challenge has been archiving and analyzing all these data, a task, he admits, that “has been like giving birth to barbed wire.”

The scientists working on the cardiac myocyte should be so lucky. Chosen as much for their esthetic appeal as its biological importance—“It’s a beautiful cell to do business with, a beautiful cell in culture, a beating cell,” Gilman notes—myocytes are as fussy as tropical fish when they’re placed in a culture dish, and they live about as long. The dilemma this team must answer is: Can the myocyte survive and, more importantly, can they survive the myocyte? Still, Myocyte Committee Chair Jim Stull is cautiously optimistic, reporting that his team can maintain the cells for at least 24 hours without death or degeneration. With a preparation that finally lives long enough to be tested, he says, they have a “green light to begin the single ligand screen for this cell type.”

At the end of the day, Henry Bourne sounds a note of caution.

Success, he argues, raises the specter of “‘magical inductionism,’ the fallacy that if you simply collect enough data, knowledge will ensue.” To “take the magic out of it,” he recommends that the group concentrate on one part of the map, rather than trying to take in the whole continent—he recommends the pathways that use PIP3 as a second messenger. And by the way, he notes, some more powerful analytic tools would be nice, too.

“You wanna see the Grassy Knoll? You come this way—I’ll take you on a tour,” offers a seedy-looking character in a dirty green T-shirt. Dinner that evening is at the infamous Texas Book Depository—on the seventh floor, not the sixth, where Lee Harvey Oswald took aim. Here, scientists collect in groups to discuss the day’s presentations, people they have known only by teleconference until today. “It’s like meeting a celebrity,” enthuses one. “You’ve only seen them on TV and then here they are in person.” To which his colleague responds, “Yeah, people looked like I thought, except they were different sizes.” Another trio is discussing the best watering holes in the Far East. Eight or nine are hogging the cheese and crackers during a lively discussion of the solubility of membrane proteins.

Asked for their reactions to the Alliance’s progress, they agree: it is like building airplanes in the sky, and, yes, imagining the looks on outsider’s faces when they’ve built their virtual cell will be worth it. One confides that most of the results were generated in the past few weeks. “It really was like giving birth to barbed wire,” he says. For members Temple Smith of Boston University and Paul Simpson of the University of California, San Francisco, however, the most impressive thing isn’t the data, but how the Alliance itself has succeeded in working as a network. “Look around the room,” says Smith. “A diverse group of scientists can overcome barriers and communicate and work together: physicists, chemists, bioinformatics specialists. Our biggest challenge now is that there are a zillion attractive things to do but not a zillion dollars, brains, or people.”

By 2003, the biggest challenge would be keeping the project from running aground.

The crisis was strictly technical, not organizational. Both of the cells chosen by the steering committee at the beginning of the project had failed to live up to expectations. “To be considered, a cell had to meet several criteria,” Gilman explains. “The most important was that we wanted to work with real cells, not cell lines. They had to be mammalian cells, from an experimental species with a sequenced genome—meaning, at the time, that it had to come from the mouse. It had to be large enough for microscopy and manipulated readily. It had to be normal—no cancerous cells, no matter how readily they took to cell culture. And it had to ‘do something.’ But we left off one criterion that should have been on the list. And that was easy.”

An observer could have foreseen the demise of the finicky, undependable cardiac myocyte; always difficult, “we were never quite certain about it,” Gilman admits. “The preparation was tortuous. We were making progress, but we weren’t sure it was worth all the effort.” The real disappointment, however, was the B cell. Able-enough performers, in the end they presented one insurmountable difficulty.

Gilman explains that everyone had agreed that the technique known as RNA interference, a way of “knocking out” proteins (singly or in combination), was a “must-have” technology. “One of our goals was to use RNA interference to disrupt pathways, even though, at the time, the technique worked only in worms and fruit flies. We were going to light candles for the people trying to do this in mammals. And they succeeded. RNA interference does work in mammalian cells—just not ours.” Despite heroic efforts, investigators could not get the technique to work in the B lymphocyte. Worse, they could not get the technique to work in their back-up option, the WEHI-231 cell—a disappointment Gilman calls the “ultimate ironic shaft.”

Reluctantly, the AfCS Steering Committee decided to give up

on primary cells, change course, and choose a cell line to be their new “gold standard.” With the 2003 annual meeting just two months away, the AfCS laboratories went to work, collecting enough data for participants to evaluate the merits of three candidate mouse cell lines: two derived from a white blood cell of the phagocyte clan, the macrophage (J774A.1 and RAW264.7) and one from the pituitary gland (AtT-20). The winner, hands down, was RAW264.7. Like most cell lines, it was easy to grow and stable in culture. It submitted willingly to RNA interference. It came from a mouse, so the tools and reagents the Alliance had already developed could be recycled. Finally, being a macrophage, “it was a performer—it secretes cytokines, it engages in phagocytosis, it moves.”

“Some might fear we wasted a lot of time,” Gilman observes. “But in fact, we learned how to do a lot of things. In particular, we had built the infrastructure.”

Actually, they built an airplane. To date, double ligand screens, involving 24 signals and four parameters—calcium, cyclic AMP, 21 phosphorylated proteins, and 18 cytokines (facilitated by the development of an assay measuring all 18 simultaneously)—have been completed. Following Henry Bourne’s suggestion, Alliance investigators limited more detailed studies of RAW264.7 signaling pathways to a single “X module” encompassing the responses of calcium and PIP3 to three ligands. In the first stage of this project, they compiled a parts list of about 200 signaling proteins. Next, using the coveted RNA interference technology, they began to examine connections between pathways. Finally, by comparing the data they have generated with “legacy” data from the literature, using state-of-the-art software, they are in the process of verifying “one of the most detailed maps of pathways leading from the activation of receptors … to stimulation of calcium”—our first glimpse of the signaling circuitry in a living cell.

A century ago signaling investigators translated the first words of the language spoken by cells. Forty years ago, they learned how to

track receptors with radioligands, and began to speak in two-word phrases. By the beginning of the new millennium they had pieced together sentences, even reconstructed portions of paragraphs. And now, four years later, they have a small and unfinished yet exquisite book to show for their efforts.

Labor Day, 2004. Gilman is working on the renewal of the glue grant that forms the major portion of the Alliance’s funding. It’s a time to review the big picture, to look back on what has been accomplished and to map out what remains to be done.

“From an experimental point of view, we’re just now getting to the fun stuff. But this social experiment is working very well. We’ve proven that we can work together and that modern communication technologies make long-range projects viable. People do have to have an appropriately selfless attitude. As far as that goes, there’s some correlation with age and, as a result, the degree of security they feel. It’s harder for younger people. But what they gain is a crack at solving a really big problem.”

“Five years isn’t a very long time,” he admits. “We can’t claim any huge victories yet. What we do see are the substantial interactions between pathways that everyone suspected were there but had never been documented. The level of complexity is huge.”

As for the next five years, Gilman is resolutely optimistic. “What can I say?” he concludes. “We’ll make a lot of progress. We have a gold mine of data, although it’s not obvious yet where all of the gold is. And in the end, we’ve generated a hypothesis machine, one that can and should inspire lots of questions.”

Ever since the publication of A Pattern Language more than 30 years ago, Christopher Alexander has felt that his theory was not quite complete, that he was missing something vitally important. Now, he says, he knows what that something was—the big picture.

In his latest work, The Phenomenon of Order, the first of four books intended to constitute an “essay on the art of building and the nature of the universe,” Alexander argues that beautiful buildings—and living things—display “wholeness,” a quality he attributes to entities he calls “centers.” He defines a center as “a physical set that occupies a certain volume in space and has a special marked coherence” and emphasizes that centers not only create the wholeness but are created by it; they seem to emerge from it as intersecting elements of a larger pattern, rather than distinct units. In contrast to the way we are used to defining space—the proverbial four walls—centers have fuzzy boundaries, are more fluid; they do not necessarily correspond to items we can name. But this imprecision is unimportant because the details of each center do not matter as much as their interrelationships, it is this “system of centers” in a building, a scene, or a community that creates wholeness.

As an example, Alexander asks readers to consider a country house, surrounded by a garden:

After careful examination, Alexander has identified 15 fundamental structural properties he believes are common to things exhib-

iting wholeness, things with “strong centers” that “have life.” Number eight is something he calls “deep interlock and ambiguity,” in which a center and its environment, or that center and another, interdigitate, forming the sort of pattern you might see in a house with an expansive wraparound porch, fine lace, or a Persian carpet. Number 15 is “not separateness.” A building with this quality melds harmoniously with the world around it, a space, such as a garden, with it is filled with things that connect to one another in some way; each part “melts into its neighbors.”

“Nature too is understandable in terms of wholeness … centers, wholes, and boundaries occur repeatedly throughout the natural world,” Alexander writes. As a result, “the fifteen properties appear as geometric features of the way that space is organized in nature.” Deep interlock and ambiguity, for example, appear in the pattern of a giraffe’s coat or the involutions of the brain. And not-separateness is an integral feature of any ecosystem, from a backyard garden to a rain forest, and “corresponds to the fact that there is no perfect isolation of any system”—even a system as small as the collective signaling pathways in a single cell.

An architect with more technical virtuosity than insight can “make buildings by stringing together patterns, in a rather loose way,” writes Alexander. A house cobbled together this way may be a shelter, but it is not a home; its walls and windows and doors are no more than an “assembly of patterns”; it stands out painfully from its surroundings. On the other hand, when patterns are integrated—for example, by locating “family space” within the kitchen, replacing a solid wall with interior windows, adding a terrace, or planting a garden—deep interlock makes the finished building “very dense; it has many meanings captured in a small space; and through this density, it becomes profound.” In contrast to the aridity of cookie-cutter suburban houses and offices composed of cubicles, the visual and emotional impact created by superimposing design elements and erasing the

boundary between the house and its surroundings adds a complexity that can transform ordinary living spaces into “buildings which are poems.”

Living things, too, have deep interlock and ambiguity, display not-separateness. The lines of communication within our cells, for example, are not just strings of proteins but inseparable interlocking patterns compressed into a space far smaller than the tiniest speck we can see with the naked eye. The exchange of chemical messengers within these networks weaves a tapestry of larger patterns that link cell to cell, tissue to tissue. From these intricate and meaningful conversations between nerve and muscle, bone and blood, mesoderm and ectoderm, cell cycle and death machinery emerge bodies that are dense with meaning, organisms that are also poems.