__Forget coyote art and adobe. Santa Fe's next claim to fame will be rescuing us from the digital data avalanche. __
This is the age of the data dump: astronomical lore by the gigabyte from the space shuttle and Hubble Telescope; DNA sequences by the billion from projects mapping the human genome; mountains of research from distant scientific outposts, government labs, universities, biotech firms, and drug companies bent on creating the next miracle molecule. Data is useless, however, until it's organized, analyzed, categorized, and understood - that is, until it's converted into information. But humans have long since proven unequal to the task of interpreting these infinite data streams. Hence the need for some grand new tools.
Today, those tools are pouring forth from an unexpected place: Santa Fe, New Mexico, home to a growing, computer-powered industry known as informatics, and the center of the data dump universe. Informatics is all about developing software that digests reams of raw data and returns cohesive information that can, for example, lead to new drug discoveries, high tech manufacturing materials, mechanisms for predicting financial markets, or methods for streamlining production supply chains. In the rarefied air at 7,000 feet, within a two-mile radius of Santa Fe's downtown plaza, roughly a dozen companies - making up what's known as the Info Mesa - are building this software, reducing the complexity of the material universe to a knowledge lode of amazing power and purpose.
Among them are outfits like Genzyme Genetics, which conducts genetic testing, and PE Informatics, which invents automated data-mining systems for the petrochemical, pharmaceutical, and agricultural industries. Phase-1 Molecular Toxicology tests and analyzes drug compounds for toxicity. The Prediction Company uses chaos theory, complex systems theory, and a raft of proprietary "advanced forecasting technologies" to predict the ups and downs of the stock market. Bioreason, Daylight Chemical Information Systems, OpenEye Scientific Software, the National Center for Genome Resources, the Santa Fe Institute, Complexica, Metaphorics, Strategic Analytics, the Swarm Corporation, and the Bios Group all offer similar services: making sense of data using a range of mathematical tools. Notably, all of these outfits - except for the National Center for Genome Resources and the Santa Fe Institute, both nonprofits - are doing something that most dot-coms haven't pulled off yet: earning profits.
At first glance, Santa Fe seems an unlikely spot for a high tech explosion - the city doesn't even have a major airport and is best known for its art galleries, New Age scene, food, and that strange mix of howling coyote kitsch and dusty-looking adobe that defines Santa Fe style. But Santa Fe has a strong technical legacy inherited from the nearby presence of the Los Alamos National Laboratory, and the same attributes of climate and lifestyle that attract tourists have also made it a desirable, and fairly affordable, place to set up a high tech business.
Granted, Santa Fe's informatics scene is not about to displace Silicon Valley as an engine of technology wealth. Informatics is still a niche industry, but it's poised for rapid growth. Robert Olan, an analyst at Hambrecht & Quist who follows the emerging field, says annual revenues for informatics probably total "a couple of hundred million," with the Santa Fe firms raking in only a portion of that. Olan points out that, at present, many informatics firms elsewhere in the country are attached to big pharmaceutical companies that are still doing their own number crunching in-house. This will change, however, as their data loads become so heavy that drug companies will be forced to outsource.
"The pharmaceutical companies spend a lot on R&D - roughly 20 percent of their budgets," says Olan. "But with an overwhelming amount of information to deal with, it has become too expensive for the big companies to look at all the data, so it's inevitable that they'll have to turn over the reins to outside informatics firms." Since the Info Mesa represents the only sizable concentration of informatics firms in the US, Olan predicts a lot of the data flow will be channeled there.
Stuart Kauffman, a professor at the Santa Fe Institute and a founding partner of Bios - which applies complexity theory to everyday business problems - says that, thus far, the Santa Fe informatics firms haven't been showered with the kind of VC funding that's flooded the Bay Area, but he believes they're on the cusp of explosive growth all the same.
"I'm told that Santa Fe feels like what Silicon Valley felt like 10 years ago," says Kauffman. Some Info Mesa entrepreneurs, including Kauffman, aren't waiting for VC money to arrive. His firm, which has annual revenues that top $4 million, is planning to go public "in one to two years." Thus far, only 1 of the 14 informatics companies on the Info Mesa is publically traded, Genzyme Genetics, which is a subsidiary of the Cambridge, Massachusetts-based Genzyme Corporation. This could change soon: Bioreason and Phase-1 are rumored to be plotting IPOs, and Kauffman expects others to consider public offerings as the Info Mesa starts showing up on the radar of more investors. All of which makes for an anomalous and exciting situation: The second-oldest city in the country - Santa Fe was founded in 1610 - is positioned to emerge as a tech-driven 21st-century hot spot.
As Kauffman points out, investors would be wise to take notice soon. With the informatics industry just finding its footing, he says, "There is still a lot of low-hanging fruit."
Anthony Rippo, a smallish, graying 58-year-old with Ben Franklin-style glasses, is the founder, chair, and CEO of Bioreason, a 2-year-old firm that uses an automated reasoning system to sift through millions of chemical samples and spot those few compounds that can be turned into drugs. With just 28 employees, Bioreason's 1999 revenue was more than $1 million, and Rippo expects to pull in $4.2 million this year.
Rippo started his first business while still in high school, where he made ribbons to sell at football games, threw ice-skating parties, and sold assorted other products and services to, as he puts it, "get a revenue stream" - something he's managed to pull off at each company he's launched.
Rippo senior was a San Diego fisherman who hoped his son would join the family business. But Anthony went to medical school, getting a degree from Loyola University in Chicago in 1966. After his internship, he started a medical practice, where, because of his family's fishing-industry connection, Rippo began getting calls from boat captains miles out at sea, radioing in with the symptoms of their sick crew members. Often enough, Rippo could diagnose ailments over the radio, but what he really wanted was a visual image of the problem. So in 1970, Rippo established Marine Medical Services, a telemedicine provider for fishermen and merchant seamen worldwide. By 1975, he was receiving video over radiotelephones from ships at sea, whose captains could now send slow-scan television pictures to Rippo's office. At the time, this was leading-edge technology.
After starting a nonprofit, an equity partner firm, nine other successful companies in the medical and sensor industries, and raising six kids, Rippo and his wife, Madeline D'Atri, were ripe for a change. They left San Diego in 1994, toured the West for six months, and wound up in Santa Fe, where they bought a 130-year-old dance hall on Canyon Road - art gallery row - and converted it into their home.
John Elling, Rippo's next-door neighbor, was an analytical chemist who worked at the Los Alamos National Laboratory, some 35 miles away. Elling also had a consulting job with Amgen, a biotech firm in Boulder, Colorado, which was spending millions of dollars on drug-discovery research and needed a way to wade through chemical compounds to identify promising new drugs. Elling told Rippo of the huge commercial possibilities involved.
"It sounded like a business to me!" Rippo recalls.
In January 1998, Rippo, Elling, and Susan Bassett, a Florida State University computer science professor who had taken a sabbatical year at Los Alamos doing machine-fault diagnosis, teamed up to start Bioreason. They hired Ruth Nutt, a retired medicinal chemist who had spent 31 years at Merck. They recruited computational chemists, software engineers, AI experts, and various other computer adepts, all of whom put their monster minds together to create an automated reasoning system that could inspect vast amounts of chemical data quickly and point the finger at potential new drug compounds. The software would examine huge databases of stored knowledge, using AI technology to compare the known to the unknown and reveal chemical relationships. This was their data-mining tool, but now they had to test it - preferably on real-world data.
Bioreason occupies most of the third floor of the Wells Fargo Building in downtown Santa Fe, a modern adobe structure a block off the plaza and across the street from the city's most upscale hotel, the Inn of the Anasazi. As is true of most Info Mesa companies, Bioreason's office space is divided among people, computer workstations, and - in a separate room, under lock and key - the servers that store the drug-compound libraries. "The crown jewels of a pharmaceutical company are its compound libraries," explains Rippo. "So security is extremely important."
The first time Rippo saw one of these libraries was in St. Louis, Missouri, while visiting Searle, a division of Pharmacia. Four months after launching Bioreason, Rippo took a tour of the pharmaceutical giant's chemical storage vaults. There, in refrigerated cabinets - miles of them, it seemed - were rank upon rank of rectangular plastic trays called well plates, each containing 96 tiny wells that hold a trace amount of a specific chemical compound. The total collection of these compounds - aka the "library" - is the source from which companies like Searle hope to discover one or more new drugs.
But as Rippo observed, finding those drugs was no small task, because Searle's library held nearly 90,000 different chemical compounds. Here was a data-dump problem of historic proportions. Still, big as it was, this was not considered an overly large library - some drug companies have close to 3 million chemical compounds in storage - but it brought home to Rippo the true scope of the problem his company had been founded to solve.
In 1999, Parke-Davis, a drug company that was collaborating with Bioreason, sent a bunch of old chemical data to the Bioreason servers. Parke-Davis had "screened" a portion of its own chemical library in 1992, meaning its scientists had determined the chemical structures of the library's various compounds and whether any of them had the potential for producing new drugs. Traditionally, it was the province of individual chemists working in "wet" labs to make this determination, and Parke-Davis' people had been probing this specific batch of chemicals for years, looking at the compounds one by one. By 1999, they decided they had pretty much exhausted the batch's potential. "The bottom line was that they felt they knew everything there was to know about that screen," says Rippo. "They had seven years' experience with it."
So Parke-Davis took the raw data its chemists had spent years analyzing and sent it to Bioreason over a secure line. With hopes that their software would discover most, if not all, of the druglike compounds the Parke-Davis chemists had found, Bioreason's scientists ran the same screen through their own data-mining systems, LeadPharmer and DataPharmer. The result surprised everyone. "In just a few hours of computing time, not only did we find everything their scientists found, we found things they had not found," says Rippo. "That was a blow-away."
Bioreason's software had identified two additional compounds in the Parke-Davis data - outliers, or singletons, compounds with potentially druglike properties that were off in some corner by themselves, and not part of a broader chemical family.
Rippo says he isn't sure what Parke-Davis did with the new data, since the company wasn't required to make its intentions known. But the pharmaceutical giant was so impressed that it's doing another collaboration with Bioreason. Since then, Rippo's company has come out with other software packages, ADMEPharmer, KnowledgePharm, and DrugPharmer; completed its second round of financing; and is contemplating a third, "or possibly a mezzanine round prior to an IPO," says Rippo, "depending on the market and our success in signing up more customers."
Twenty years ago, none of this would have been possible: The computer power didn't exist, the software didn't exist, and the vast chemical libraries didn't exist. Then again, neither did another necessary element of the overall equation: a language by which chemical data could be entered efficiently into a computer - as easily as a line of text - and transmitted over phone lines. When Parke-Davis sent its chemical screen to Bioreason, however, the data was expressed in just such a language - known as Smiles.
Smiles is the brainchild of longtime Info Mesa stalwart Dave Weininger of Daylight Chemical, a company that does rapid analyses of massive chemical databases and earns more than $4 million annually. Smiles, a sort of universal language for chemical compounds, is an acronym for simplified molecular input line entry specification. But it's also called Smiles because that's what chemists did whenever Weininger explained his system to them. He'd tell them how it's both universal and independent of any natural language, how it's able to express the entire breadth of chemical compounds, and how its formulas can be fed into any computer by following four simple rules. They'd listen to all this, nod their heads, and smile incredulously.
They had good reason for skepticism: Until Weininger invented Smiles in 1983, chemical compounds were represented in any one of three ways, none of which were both universal and computer-friendly.
To begin with, there's the systematic name of a compound, the way it would be designated in natural language: acetylsalicylic acid, which is aspirin, or dimethyl ketone, otherwise known as acetone. Such nomenclature was good enough for simple compounds, but consider the systematic names for slightly more complex ones, such as 3-(para-hydroxyphenyl)-2-butanone, 2-methoxy-5-methylpyrazine, or thiopropionaldehyde-S-oxide. Then consider that in, say, Dutch, German, or Japanese those names don't transliterate, but instead are often derived from different words altogether.
The second way of identifying a chemical compound is by its molecular formula, which can be simple - H2O (water), NaCl (salt), or H2SO4 (sulfuric acid) - or complex, as in O2CC6H4CO2C2H4, which is known by the trade name Dacron in the United States, Trevira in Germany, and both Terylene and Crimplene in the UK. The molecular formula also masks a hidden ambiguity. Though it names the elements present in the compound and their relative abundances, it is silent as to the compound's molecular structure, which means that one formula can apply to two or more substances depending on how its elements are physically arranged: C2H6O, for instance, is both ethanol (with the three elements bonded in one structural arrangement) and dimethyl ether (in which the same three elements are arranged differently).
These problems and ambiguities are eliminated in the third conventional way of representing a chemical compound - with a diagram that indicates its precise molecular configuration. Water, for example, is:
O / \ H HAlthough that diagram will look the same to chemists everywhere, there is no easy way to enter lists of such pictorial structures in a searchable computer database, especially when the molecules themselves, and the resulting diagrams, begin to get complicated.
So the problem was that chemistry, the world's most practical, necessary, and all-pervading science, had no universal, language-independent, computer-parsable nomenclature in which to express itself.
Then Dave Weininger invented Smiles. Weininger is from Schenectady, New York, where his father worked as a chemist and passed to his son a passion for the science. Growing up, Weininger's hero was Emil Fischer, a German experimental chemist who discovered the chiral nature of sugars, for which he won the second-ever Nobel Prize for chemistry, in 1902. Weininger was so interested in how Fischer's mind worked that he translated the chemist's biography into English from German. By the time he left for college, Weininger knew chemistry the way other kids know baseball.
In the mid-1980s, after earning a PhD in civil and environmental engineering at the University of Wisconsin, Weininger got a job with the EPA. Dismayed by the number of toxic chemicals in the environment, Weininger wanted to help get rid of them. His work required that he enter the names of countless toxic chemicals covered by the Clean Water Act, the Toxic Substance Control Act, and other protective measures into a computer database. He soon found himself drowning in a sea of chemical nomenclature that, even to him, proved baffling. So, as a quick-and-dirty shorthand, essentially for his own personal use, he came up with a chemical notation system governed by four rules:
The rules proved able to represent a large class of organic compounds in a way that could easily be fed into a computer. Acetic acid became CC(=O)O, which could be written as a line item in a computer by anyone who could type. After adding a few symbols to cover more complicated matters (like isomerism or chirality), Weininger decided he had invented a truly universal, computer-parsable chemical notation, one in which, as he later put it, "an Australian chemist in 2025 will be able to understand a Smiles generated by a Japanese chemist in 1985. There's no assumption that they share common computer software, hardware, and so on."
In 1987, Weininger incorporated Daylight Chemical Information Systems. The company signed licensing agreements with several of the world's major chemical, drug, and agricultural companies, as well as some government organizations, and he was soon making extremely good money - with practically no overhead. Within five years, Daylight's annual profits topped $1 million.
Weininger and his small crew of chemo-informatics hackers went on to produce other specialized software packages for the working chemist, including Rubicon, which is a rule-based geometry program for creating 3-D forms, and Thor, a client-server database for chemical information.
Merlin, the Daylight search engine, runs through a database of millions of chemical compounds in a matter of seconds every time it answers a question. Even Weininger himself is sometimes surprised at Merlin's power. He likes to recollect the time he had a demo running at an American Chemical Society meeting, and a guy came up to the display table and asked, "Are there any Japanese patents for Best?"
Weininger asked, "Best?" He'd never heard of it.
He typed in the word, hit Enter, and in seconds Merlin returned the news that Best was the trade name in Argentina for a compound known as diazepam (C16H13ON2Cl, better known as Valium in the US), for which a Japanese patent had indeed been issued. The Merlin search also returned the compound's molecular structure and listed its chemical reactivity, sorted by a choice of several parameters selectable by the user, along with the by-products of those reactions, and other chemical minutiae.
Such marvels aside, the company's real moneymaker is the Daylight Reaction Toolkit, a magical system that allows the user to select bunches of chemical compounds and have them "react" together in a virtual chem lab. This is a revolutionary tool because it's a way of doing chemistry without actually performing the experiments: The computer does them, predicting their outcomes on the basis of the various reagents' known properties, all of which are stored in memory.
"A chemist using this system can do a million experiments on Monday," says Weininger. "If he's unsatisfied with the result, he can do a million experiments the next day, go back and hit the three that look promising, then actually do the wet chemistry on Wednesday, and then have it written up that weekend."
This is reducing chemistry to information, a subject Weininger speaks about in somewhat messianic terms. "If you want to get the tools to do a traditional wet experiment, there are a hundred companies that will sell you the stuff - Bunsen burners, test tubes, compounds, connectors, and so on. If you want to do chemistry as an information science, there's only Daylight. We don't build the black boxes that do the work, we build the things that draw the pictures, canonicalize the names, look them up against other people's data, and let you publish your data so others can understand it."
Today, Daylight has more than 250 corporate clients and sells its software for anywhere from $10,000 to $250,000, depending on how a client plans to use it. But the prevalence of Smiles is even more widespread. "Almost 100 percent of pharmacological, agrochemical, even patent companies are using some form of our product," says Weininger. By stuffing the modern chem lab into silicon chips, Daylight has brought us the era of clean-hands chemistry.
Drug discovery without drugs, chemistry without chemicals. Suddenly, everything was information.
When it became clear to Santa Fe scientists that a New Mexican version of Silicon Valley was springing up in their midst, they decided to name their city accordingly, with a memorable phrase to capture its essence. One idea was Silicon Arroyo, but that was too copycat. Another was Data Mountain, which was not half bad. A friend of Weininger's came up with the name that stuck: Info Mesa.
There are good reasons why these Info Mesa marvels have been happening in Santa Fe rather than, say, Lubbock, Shreveport, or Chicago. They can be traced back to J. Robert Oppenheimer, the physicist who in 1943 selected Los Alamos, the site of a boarding school in the Jemez Mountains some 35 miles west of Santa Fe, as scientific headquarters for the Manhattan Project. The area was chosen for its isolation and natural facilities, but the Oppenheimer family ranch also happened to be located nearby, in northern New Mexico's Pecos Wilderness. A crew of physicists that included Oppenheimer, Edward Teller, Enrico Fermi, and Richard Feynman arrived at the secret mountaintop lab and proceeded to invent the atomic bomb. Later, as the Los Alamos lab developed the hydrogen bomb, it relied increasingly on supercomputers to calculate the paths of explosive shock waves and the effects of other nonlinear phenomena - events that, as a result of their complexity, weren't easily construed by the differential equations of traditional Newtonian mechanics.
In later years, with a dwindling number of new bombs in development, the Los Alamos National Laboratory was faced with a massive oversupply of supercomputers and PhDs. It was during this era, in the early 1980s, that George Cowan, a former director of research at Los Alamos, had the idea of founding an interdisciplinary research center in Santa Fe. The center's scientists would systematically address just the sorts of problems that could be modeled with the aid of massive computation, generic problems they were already good at handling, like turbulent fluid flow, weather forecasting, neural communication patterns in the brain - plus more complicated ones, like the evolution of biological diversity within ecosystems and stock market behavior.
In 1984, with grants from the Department of Energy, the National Science Foundation, and the MacArthur Foundation, the Santa Fe Institute opened for business, with Cowan as president. It was situated on Canyon Road, in a low adobe building that had once been a convent. Murray Gell-Mann, the Nobel Prize-winning physicist who owned a home in Tesuque, just north of Santa Fe, and who had recently taken an interest in what he called "complex adaptive systems," came on as chair.
The place was soon awash in the latest scientific buzzwords: emergent behavior, autocatalytic networks, self-organization, cellular automata, genetic algorithms, ecological dynamics, artificial life, collective intelligence, chaos, complexity, the physics of information.
Indeed, information seemed to be the one constant underlying all the wildly divergent phenomena that the institute's members modeled on their Sun workstations. The lock-and-key fit between a protein molecule and a cell receptor, the nerve impulses transmitted between neurons, and the price signals sent by buyers and sellers in the marketplace are all various kinds of information. Even the physical forces that portions of matter impart to one another in the process of fluid flow could be viewed as bits of tangible information. Now, all the business of the natural world seemed to be transacted by means of information, whose complex patterns could be modeled on the computer.
As the commercial possibilities for this type of analysis became apparent, it was the Los Alamos scientists who rushed in to exploit them. Bioreason's three cofounders - Anthony Rippo, John Elling, and Susan Bassett - were Los Alamos escapees, as were the founding mothers and fathers of several Info Mesa enterprises - the Prediction Company and Complexica, among others. In many cases, even the technical staff and administrative assistants at these outfits hailed from Los Alamos, the Santa Fe Institute, or both.
The star of the Santa Fe Institute, possibly exceeding in creativity even the fabled Gell-Mann, is Stuart Kauffman. He has spent 14 years at the Institute doing computer modeling of genetic regulatory networks, thinking about the origins of life, and trying to comprehend the complexity of the natural world. Along with Cowan and Gell-Mann, he is one of the originators of modern complexity theory.
Kauffman has had one of the more exotic careers in recent science: A philosophy major at Dartmouth and Oxford, he went on to get a medical degree at the University of California. "I figured somewhere I had to learn a bunch of facts, and if I went to medical school, the bastards would make me learn a lot of facts, and that's exactly what happened." He practiced medicine, doing an internship at Cincinnati General Hospital for all of a year. Then he switched to theory.
Kauffman studied fruit fly genetics, as well as cell differentiation and development at the University of Chicago; molecular evolution and combinatorial chemistry at the University of Pennsylvania; and then, as a professor at the Santa Fe Institute, he branched out into even more abstract theory, studying the myriad ways that complex systems self-organize and operate. This was a logical progression, since all the systems he'd been studying earlier were ones in which a multiplicity of components interacted with one another to produce various outcomes, a description that virtually defines complexity theory - the science of how wholes arise out of complex, mutually interacting parts. Kauffman's work encompassed much of what companies on the Info Mesa specialize in today - looking at many disparate points of data and hunting for meaningful relationships within the numbers.
Along the way, Kauffman got a patent on a new method for making several varieties of organic molecules (a technique for creating the kinds of chemical libraries used by pharmaceutical companies in their search for new drugs), licensed it to Applied Molecular Evolution, and started collecting substantial royalties. Like other Info Mesa scientists, Kauffman isn't bashful about turning theory into money and estimates that, not counting his Santa Fe Institute earnings, he has made more than $1 million a year from royalties and consulting fees.
In 1995, a member of the Boston consulting firm Ernst & Young approached Kauffman with a business proposition after reading his book At Home in the Universe, a dense text that draws parallels between coevolution, markets, and corporations. Out of this came the Bios Group, located on Paseo de Peralta, Santa Fe's inner loop. The company, now some 70 staffers strong, advertises itself as supplying "adaptive solutions to complex business problems," which means it applies complexity theory to commerce and industry.
One of the firm's clients was Procter & Gamble, which came to Kauffman in 1998 with a problem involving its supply chain. P&G is a $38 billion corporation that controls and consumes a great many assets and raw materials, processes them along parallel and intersecting pathways, and produces a large variety of wares that it then distributes all over the world. At one point, some senior managers wondered if their "earth-to-earth" supply chain - the long trail of resource allocation, manufacturing, distribution, and customer consumption - might not be streamlined somehow. Even an incremental increase in its overall supply-chain efficiency, they knew, could yield enormous savings and higher profits.
But this was a problem P&G wasn't capable of addressing itself because, paradoxically, it didn't know what its own supply chain was - not conceptually, at least. The company was responsible for it, operated it, oversaw it, and ran it, but didn't understand it on a theoretical level. If anyone in the world could figure it out, P&G decided, Stu Kauffman could.
Kauffman and his team conducted a full study of P&G's supply chain. It was characterized by three major parameters: total inventory in the system; total time in the system; and out-of-stocks on the shelves. Of these, the only one that couldn't be fiddled with was out-of-stocks. Without exception, P&G wanted to have Tide, Comet, and the rest of its product lines on the shelves at all times.
The Bios scientists eventually produced five models of the P&G supply chain and ran them on their workstations thousands of times under different settings and conditions, creating, in Kauffman's words, a "policy space with lots of knobs you can tune." (Kauffman is a fearless coiner of rhetoric.) As the scientists observed the results, they noticed that one particular effect kept cropping up, the appearance of what Kauffman calls "lumpy integer constraints."
A lumpy integer constraint is a requirement that a given input or output must be expressed in whole numbers. P&G had unwittingly introduced such a constraint into its supply chain by imposing a general mandate on its cargo trucks, requiring all shipments to be made in full truckloads only; partial loads weren't permitted. Such a requirement makes obvious and intuitive sense. Having your trucks full when they leave the loading dock maximizes their utility and efficiency, leaves no wasted space, saves diesel fuel, reduces air pollution, and minimizes duplication of effort.
What the Bios Group's simulations found, however, was that adherence to the full-trucks rule caused disruptions elsewhere in the system. It converted smooth, or laminar, flow into an irregular and jagged shipping stream, creating bottlenecks - and even temporary out-of-stocks - as trucks waited for their cargo holds to be filled. Relaxing the full-trucks requirement would iron out any and all supply-chain kinks.
"We discovered," says Kauffman, "that if you softened the integer constraints just slightly so that you could send less-than-full truckloads, you stabilized laminar flow."
Today, some of the world's biggest and most highly visible companies are paying Kauffman for advice and guidance, parting with big cash sums in the process. Kauffman says Bios revenue for 1999 was $4.8 million and has been doubling every year; the Bios client list now includes Boeing, Texas Instruments, Unilever, Honda, and Johnson & Johnson.
Kauffman recently got a call from the Joint Chiefs of Staff, who wanted Bios' help with a problem concerning the sudden tactical changes that are often made on the battlefield: how to switch the assault from, say, Hill 19 to Hill 20. Kauffman applied the same analytical tools, and found that there, too, by softening the lumpy integer constraints just slightly, the army could "gracefully deform," as he puts it, and take the new hill.
"This is the same kind of problem as that of introducing flexibility into the P&G supply chain,"he says. "If one way of doing things is blocked, there's a way around it, and you don't get stuck."
Military strategy, product distribution, genetic regulatory systems - to Kauffman, they're all assorted items in complexity space.
"The sciences of complexity," he says, "are going to be the sciences of the everyday world."
Every Friday, starting a little before noon, Dave Weininger hosts a group lunch at Daylight's research headquarters on Route 285, just south of the Radisson Hotel. The L-shaped, three-story building is set artfully into a hillside. On the top floor, Weininger's office looks out over Santa Fe and, beyond it, the Sangre de Cristo mountains. On the ground below his three-pane picture window is a towering sculpture, a dozen or so color-coded metal spheres held together by beefy steel tubes, all of it representing the molecular structure of an experimental cognition-enhancing drug.
In the months since he inaugurated the Friday group lunches, practically every Info Mesan has started showing up: Anthony Rippo, John Elling, Susan Bassett, and the rest of the Bioreason team; Stu Kauffman, Christine McLorrain, and other members of the Bios Group; Roger Jones, CEO and chief scientist at Complexica; and the whole crew from the National Center for Genome Resources. On any given Friday, about 20 to 30 people gather, nibbling on pizza, salads, cold cuts, and raw vegetables, and drinking sodas, cappuccino, or one of three brands of mineral water. Later, after sorbet and cookies, Dave Weininger offers his patented tour of Daylight's physical plant, including visits to the famed molecular sculpture, the secure server area, and the "rumble room," where, twice a year, the company's chemo-informatics hackers get the chance to explain why their latest innovations should be included in the next software release.
Anthony Nicholls of OpenEye is a regular at the Daylight group lunch. Nicholls, a biophysicist and one of the up-and-comers on the scene, first came to Santa Fe in the summer of 1987 to attend the Matrix of Biological Knowledge conference - "wonderful name, wonderful conference," he recalls, "bioinformatics before that unfortunate word was coined."
Nicholls is from Plymouth, England, the primeval home of rain, gloom, and fog, so he was thrown for a loop by the crisp air, blue skies, and unlimited visibility of northern New Mexico. "As an Englishman," he says, "you grow up in a very claustrophobic kind of country, and then you come out here and you're able to see 200 miles!" He decided during the five weeks of the Matrix conference that if he ever had the option of living anywhere in the world he wanted, Santa Fe would be the place.
In 1990, during his postdoc at Columbia University, Nicholls developed a program called DelPhi, which worked out the electrostatic potentials of protein molecules. The program was handy, but it took an hour or more of computing time to calculate a response, so Nicholls decided he would try to speed it up. After a few months of rewriting the code, the software ran 60 times faster, divulging an answer in a minute or so. The optimized program, DelPhi II, marketed by Biosym (now a part of Pharmacopoeia), is today a mainstay of biophysics. Nicholls began collecting royalties on the software, receiving a check each February for a few thousand dollars. He deposited the checks into a savings account and forgot about them.
Next, Nicholls focused on creating a new software system that, in a matter of seconds, generated a 3-D picture of a protein molecule's surface structure. This three-dimensional view is important because molecular reactivity is largely a lock-and-key phenomenon - a small molecule fits into a concave part of a large molecule and blocks its action, for example - and the ability to visualize a protein's surface would be an unparalleled boon for biochemists.
Nicholls called his new program Grasp, for graphical representation and analysis of surface properties, and it was so popular with protein scientists that it quickly became the default system for depicting the exterior structure of any new protein. Today, whenever a protein molecule is pictured in Science, Nature, or other scientific journals, it has almost invariably been produced in Grasp. All such illustrations use the red, white, and blue color scheme that Nicholls adopted largely because he is red-green color-blind.
Nicholls first met Dave Weininger while demonstrating Grasp at a conference in Albuquerque. The two minds shared a similar wavelength, and it wasn't long before Weininger had persuaded Nicholls to quit the sheltered academic world and move to Santa Fe, where he set himself up as an independent scientist. In 1996, with the money he'd saved from DelPhi II royalties, Nicholls left Columbia, moved west, and founded OpenEye.
Company headquarters is the living room of Nicholls' three-room apartment on a dirt street so small and obscure that even FedEx drivers have been known to call for directions. Here, working on a mess of computers in the shadow of a potted plant, Nicholls is producing his dream software, a system that will do for small chemical molecules what his Grasp program did for proteins. If it works as planned, the new program will let a researcher type in a Smiles for any given chemical compound and the program will respond instantly with full-color, 3-D portraits of similarly structured molecules. For the first time, medicinal chemists, drug discoverers, and other researchers would be able to see the physical contours - and will therefore be able to estimate the chemical activity - of new and unorthodox molecular structures.
The usefulness of this software can be judged by the fact that three of the world's top chem/pharma companies - Glaxo Wellcome, Vertex, and Zeneca - have conferred generous cash sums upon Nicholls in return for the privilege of acquiring the OpenEye software, even before there's a finished product. An interim program, ZAP, is already up and running, and will soon be available to commercial users.
Six months ago, OpenEye's corporate motto was "Software just like Mom used to make." Today, it's "Kicking maximum ass." Nicholls, who isn't given to macho posturing, is sanguine about the prospects of his developing product line. His ultimate goals, he contends, are to contribute to science and to help people. "We're not just going to shave 1 percent off making a car for GM," declares Nicholls. "We're going to make something that actually affects people's lives."
The night of the total lunar eclipse last January, Dave Weininger held an "eclipse party" at his home, inviting a handful of other Info Mesans. He lives with his partner Dawn Abriel, an emergency medicine physician, in a large house on Stagecoach Road in the hills north of the city. The neighborhood, called Hidden Valley, boasts its own miniature Stonehenge, a replica of the original. Weininger and Abriel's house has more than its share of computer artifacts, with iMacs scattered wherever you might conceivably need one, plus a home office crammed with Dave's extended collection of assorted monitors, servers, and other odds and ends. The house belonged to science fiction writer Roger Zelazny until his death, and Weininger's office occupies the room where Zelazny wrote his novels.
Daylight has made Weininger and his associates rich men - not that this is any distinction in Santa Fe, and particularly not with the exponential growth of the roughly dozen data-crunching firms calling this area home. Still, the boom has made Weininger a killing in the suddenly lucrative informatics industry, and he has an impressive collection of toys to show for it.
There is, for example, his Alon A-2 Aircoupe, a small single-engine plane he has owned since the 1980s. He once landed the craft on Monterey Boulevard in Highland Park, California, after its engine quit over Dodger Stadium. Though the Aircoupe doesn't sport a bumper sticker that reads MY OTHER PLANE IS A BOMBER, it could, because Dave also owns a BAC Jet Provost TSA, a British military trainer. Yes, that's right, his own personal jet fighter-bomber.
His most recent acquisition is an astronomical observatory, which he set up in his backyard on 30 cubic meters of concrete. It's a comprehensive installation, complete with a 16-inch Meade LX-200 reflecting, motorized telescope surrounded by a movable dome - just like Mount Palomar - all of it controlled by its own dedicated Strawberry iMac, which searches for a stellar target and then reports, by voice, "Object found." (All these playthings and more are pictured on Weininger's homepage: www.daylight.com/~dave.)
Weininger and some other Info Mesans are out there now, under the dome, watching the moon as it slips into Earth's shadow. Later, they will come inside, ruminate on what could become the next wave of informatics innovations, and then get to more serious business: reading poetry to one another in the big video room, while color pictures of Earth, as viewed from the space shuttle, pass by on the wall-size video screen - slow-motion wallpaper in the background.
It's dark, clear, and cold here on the outskirts of the Info Mesa. Above, the lunar disk darkens and takes on a reddish tinge. Off in the distance, unseen, are those mythic icons of Santa Fe style - a few lone coyotes, howling at the moon.
