Bioinformatics: Leveraging for Dollars

Selling research software is a tough way to generate big returns: there are dozens of competitors, most markets are small, and technology changes fast. Some companies are trying to create one-stop shopping solutions for large clients, building consulting and training businesses onto software sales in order to leverage up per-client fees. Others are creating one-stop-shops of a more democratic kind by exploiting the Web. Instead of selling to bioinformaticians, they're aggregating often easy-to-use versions of bioinformatics software and selling affordable subscriptions. But their real revenue generation model may be context-sensitive e-procurement: pointing out to customers which reagents to buy and then taking a percentage of the selling price.

Selling research software is a tough way to generate big returns: the market is small and technology changes fast. To find profits, players are either going very deep into a few clients—or using the Internet to transform what they sell.

by Roger Longman

• The specialist market for bioinformatics software has so far proven relatively small; there are few specialists to sell to and their needs are far from uniform.

• A number of companies figure the best way to generate substantial profits is to create one-stop shopping solutions for clients, building consulting and training businesses onto software sales in order to leverage up per-client fees.

• Others are hoping to generate high per-client revenues by renting out infrastructure. The strategy is largely a play for the business of smaller companies since large firms aren't likely to rely on outside vendors, and particularly start-ups, to run key internal information sources.

• The Web has also created new bioinformatics markets: instead of selling to bioinformaticians, several players are offering easy-to-use versions of complex software to the much broader market of life scientists.

• To further increase their revenue potential, these Internet bioinformatics businesses are edging into lab reagent e-commerce, using context-sensitive selling to move in on territory pioneered by SciQuest and Chemdex.

In 1996, three American bioinformatics start-ups were funded by different groups of important venture or corporate backers—NetGenics Inc. , Molecular Applications Group (MAG), and Pangea Systems. The need was simple: turn the genomic, chemical, and screening data pouring out of the various high-speed discovery systems into usable information.

And the deal economics seemed compelling. If drug companies were willing to pay millions of dollars for combinatorial chemistry services, subscriptions to genomics databases, and high-throughput screening systems, they'd certainly be willing to pay similar amounts for the software that enabled them to use all that information efficiently.

Four years later, the need is just as acute. But of the three companies, only NetGenics continues to work more or less along the same lines on which it was founded—though it is still unprofitable and has just a handful of clients. MAG is out of business, most of its technologies sold to Celera Genomics Group , a unit of PE Corp. [See Deal], and Affymetrix Inc. Pangea has metamorphosed into DoubleTwist Inc. , a would-be life sciences Internet portal. And dozens of new competitors, pure-plays and diversifiers, have streamed into the business.

In part, the obstacles to success in bioinformatics have been created by the industry's very amorphousness. It's difficult to define the business space or the customer. Customer needs evolve; competitors appear suddenly and from unexpected quarters. InGenuity Systems Inc. , a start-up funded by Accell Partners, is developing powerful search technology created by a group of computer scientists out of Stanford University . Spotfire Inc. has a popular user interface which it's selling not merely to life scientists but to other industry researchers as well. As eBioinformatics Inc. and DoubleTwist move into the e-commerce business, selling research tools as part of their mass-market approach to bioinformatics, they compete with players in apparently different segments. For example, they'll run into BioSpace.com Inc.—a portal offering a wide range of free data and building its business on selling a variety of things like job ads, conference listings and research reagents—and Chemdex Corp. (DoubleTwist in fact just signed a distribution agreement with Chemdex's arch-rival, SciQuest.com Inc.)

To be sure, NetGenics, MAG, and Pangea were, to a greater or lesser degree, blindsided by the arrival of the Internet as a mass market communications tool. But these companies have always been cognizant of the even bigger challenges they face: the difficulties of convincing many different organizations to standardize on particular software tools; their inability to provide all the required tools; and finding enough buyers willing to pay a sufficient amount of money to make the software business profitable.

Indeed, the key problem remains that the market for bioinformatics, at least as originally conceived, isn't large enough to support a company built primarily around one or two high-cost software programs targeted to a relatively small user group. There are just a few thousand bioinformaticians worldwide and many of those write their own applications. Nor are they, or their employers, a homogeneous group: they pursue different scientific specialties with different software needs. Purveyors of bioinformatics have thus found it tough to find broad audiences for enough software applications to justify their businesses.

But for all of the apparent resistance to the kind of high-profit deals life-science investors like to see, everyone believes the bioinformatics market could be huge. Dan McCurdy, VP, life sciences for International Business Machines Corp.estimates drug companies are spending $1.2-$1.8 billion externally on R&D software and IT equipment. "As pharmacogenomics and molecular medicine expand," he says, "the opportunity over time—particularly as clinics and providers move into individualized medicine—expands with them into the many tens of billions. And that's not including the markets for chips from companies like Affymetrix and Motorola."

IBM wants in, figuring that its current businesses will both support and profit from a fling into life sciences. It announced that its Gene Blueproject—the company aims to build a computer 500 times more powerful than the fastest computers being used today—would begin with an attempt to solve the problem of protein folding, a key challenge for drug makers trying to investigate the behavior of drug targets. Nearer term, the company is applying database technologies it's developed for linking sets of disparate kinds of data into its Discovery Link software, which it hopes will provide much of the software infrastructure for bioinformatics applications developed by more specialized firms, as well as a way into drug companies for IBM's other businesses.

A number of bioinformatics start-ups have begun to think the same way: use software to wedge their way into client companies and then dramatically expand the business with consulting-type or infrastructure-based services. Companies are pursuing a number of strategies for doing so, but the point is the same: solve a big bioinformatics problem, not a small one. NetGenics' CEO Manuel Glynias figures that the top 40 drug companies will each spend $5-10 million annually on "outsourced bioinformatics," particularly as the industry begins to adhere to standards. With that kind of market, NetGenics can build what Glynias calls "a reasonable public company."

But four years into the bioinformatics industry, only Lion Bioscience AG among the start-ups has turned the trick in a Big Pharma client, with a stunning $50-100 million bioinformatics contract with Bayer AG [See Deal]. Other start-ups, recognizing the obstacles among large firms, are targeting smaller drug companies and biotechs, looking to provide a rented bioinformatics infrastructure for players who can't afford to build their own.

On the other side of the world are those companies attempting to bring bioinformatics to the masses using the Internet—a tool whose power they barely understood just three years ago. Compugen Ltd. and DoubleTwist have both democratized their specialist offerings with easier-to-use Internet versions; eBioinformatics, founded this year, is doing much the same thing.

Still largely unheard from are those companies who want to use bioinformatics tools to keep clients addicted to their proprietary data—Celera in particular. But if each of these companies is following different paths, and to some extent different customer groups, they all seem to want to do one thing: provide a one-stop bioinformatics solution.

This is easier said than done. With few barriers to entry (creating a new software program can cost just a salary or two), the field has dozens of for-profit and academic competitors who are continually introducing software refinements which can obsolete key current applications. Keeping up with this deluge of innovation is difficult, if not impossible. Many of the smaller players don't have venture backing and thus don't have the same growth pressures as the better-known start-up; these boutiques can focus on doing one thing well, even if the opportunity provides little more than good jobs for the employees.

Along with the boutiques nipping at their heels, the bioinformatics shopping-centers will all be competing with each other—despite what they say are unique niches. NetGenics' proposed corporate clients are considerably upmarket, in terms of intended per company spending, to the vast majority of DoubleTwist's Webbed-targets, but both are effectively trying to bring their customers a wide choice of informatics tools for combining the disciplines of genomics, chemistry, and screening. IBM wants to use Discovery Linkas the database architecture all bioinformatics programs will use; Lion has the same ambition for its SRS software (and far more bioinformatics clients), though it also wants to provide a broader bioinformatics solution than does IBM. Viaken Systems Inc. aims to partner with a variety of bioinformatics suppliers so that it can provide outsourced infrastructure to mid-sized companies—but there's no reason to believe that it won't try to sell the same thing to larger companies, in direct competition with NetGenics and Lion.

The Integrators

If the bioinformatics industry can point to one big success, it is Lion BioScience: the only pure-play to have won both a relatively broad customer base as well as a client big enough to impress even an IBM.

It's also one of the few major bioinformatics start-ups without a big financial backer. Lion specifically rejected venture funding, which was by no means lacking in Germany. Its executives wanted a chance to explore the intersection of biological research and information technologies, without the pressure VCs would impose to focus on a quickly IPO-able story. "We believed," notes Claus Kremoser, Lion's VP, corporate development, "that it was necessary to integrate new information technology with the genomics approach [to drug discovery]. But doing so is complicated and requires time to build. It's like a moon landing: it needs to be done in stages."

Lion's launching pad concept was to create a relatively low-cost entry that would subsidize the business and simultaneously introduce the company to its core customer base. It began, therefore, with a contract sequencing business—relatively rare in Germany when it began, but now with plenty of competitors who have been driving down the $3.5 million business' margins, says Lion's CEO Friederich von Bohlen.

The company's original bioinformatics software was adapted from academic shareware for protein structure prediction and function annotation written by two Lion co-founders then at the European Molecular Biology Laboratory . Lion took a license to the program, Gene Quiz, and developed it into the more robust and functional BioScout.

Lion likewise adapted its second and more popular package, SRS(for Sequence Retrieval System) from an academic precursor package. The program hunts for information related to any specific genetic sequence by searching through more than 400 different databases—sequence, genotype, medical genetics, and literature databases. A license to the program, bought by about 18 companies so far, including, says Kremoser, "most of the major ones," costs the average client around $60,000. BioScout has about seven customers, paying on average about $200,000.

But with "double-digit millions" now in revenues, says Kremoser, the aim of the company is merely to sell its software as a way into its clients. Indeed, von Bohlen notes that while the company's average license income from software sales is about $100,000, the average Lion client spends about $500,000 because they need additional customization and support. Lion has SRSdevelopment programs, he says, with even the most sophisticated companies, like Glaxo Wellcome PLC and SmithKline Beecham , "because they need the support."

"The classical business model in bioinformatics—developing software and selling licenses—is not sufficient," says Kremoser. "In a universe of 50 big drug companies there's not enough business in selling $100,000 licenses. So you either need to extend the borders into other industries, like Spotfire, which sells to drug, chemical companies, etc., or you need to increase the value of each client into the tens of millions of dollars. And there the universe of 50 companies is enough."

Lion's goal is to enable, within its chosen sphere of data accessing and analysis, a complete corporate bioinformatics solution. Such applications will go far beyond what Lion can develop on its own and far beyond biology. It's already added to its genomic software a plug in called ArrayScout, for integrating expression profiling data from chip arrays into the broader sets of gene information available from BioScoutand SRS. But Lion now wants to bridge over into the chemistry side of the research house, effectively allowing researchers to start with a target, add data from various experiments like gene knockouts and animal models, annotating as much information about screening hits as possible, and determining the kinds of compounds likely to work.

Beyond the Bioinformatics Sophisticates

So far, Lion has developed a prototype biology-chemistry bridge called GenChem, working with an unnamed US chemoinformatics software company. "We'll never do everything—biology and chemistry and patient information and pharmacogenomics—on our own," according to Kremoser. Lion, he says, wants to be the integrator of bioinformatics applications, recognizing that it can't be expert in everything along the drug discovery process. "If we need another partner on advanced search technologies or on database technologies, we'll invite them to participate. But we'll provide the pharmaceutical company one reliable partner."

For a few bioinformatics sophisticates—SmithKline Beecham, Glaxo Wellcome, and Merck & Co. Inc. , in particular—Lion may not have much to offer beyond its current software packages and a few it sources and distributes. With up to 100 or more bioinformaticians, these companies can design de novoor adapt from other off-the-shelf programs specific applications to meet their own needs and already have the infrastructure in place.

But the majority of the top 20 need plenty of bioinformatics help—a premise on which Lion's business model is based. And Lion's aim is to step them up from one of its relatively inexpensive tools to an entire multi-million dollar package of research informatics applications and services, providing "one-stop shopping for bioinformatics—the general contractor," says Kremoser.

And one reliable bioinformatics partner is what customers want, he contends—arguing, however, from a very small sample size: Bayer, Lion's one major client. "Bayer had 70 software packages," he says, few of which could communicate with each other. Instead of speeding up the process of discovery, they were slowing it down "because they had created isolated islands of research." What's required, he says, "is for the preclinical researcher to have access to screening data or to data from expression profiling experiments. To have this seamless system you need one company to talk to, one central integrator."

Lion's Bayer deal, probably the largest bioinformatics collaboration signed to date, sets up a Cambridge, MA-based Lion subsidiary called Lion Bioscience Research Inc., dedicated to Bayer's bioinformatics needs. Bayer is paying an estimated $6-7 million annually for 4-5 years of FTE support, plus up-front and license fees, milestones, and an equity investment that got it 11% of the company—all told perhaps $50-60 million in more-or-less guaranteed funding. Bayer gets exclusive rights to the resulting software for 12 months, after which Lion is free to sell it to all comers.

Bayer wants Lion to build, using the SRS/BioScoutplatform, a set of applications to enable the company's genomicists, screeners, chemists, pharmacologists and toxicologists to pre-select the best target candidates based on in silico information. Gunnar Weikert, MD, Bayer's worldwide director for what it internally calls "physiomics" (a handy rubric for genomics, proteomics, pharmacogenomics and bioinformatics), likens it to building Word, Excel and PowerPoint onto a Windows Office platform.

For example, LBRI will build, probably working off the GenChemapplication, a system for capturing the screening data Bayer scientists are generating. This data is often lost in individual databases constructed by specific research groups, notes Kremoser, so that when the libraries are re-screened, much of the information gleaned from earlier screenings has to be regenerated.

LBRI will also create a worldwide communications system, linking Bayer's seven research sites. The sites are now organized primarily by disease area. But since, at least in the earliest stages of research, it's difficult to tell which targets are likely to work in which diseases, the data these sites generate need to be easily available to researchers elsewhere. Lion is thus developing an intranet for connecting all research sites through the central bioinformatics system in Cambridge, which will itself be screening a variety of databases over and over again for genes and targets specified by the research groups. Once LBRI identifies a new gene candidate, it will be delivered to the relevant clinical group.

The whole process will require LBRI to do substantial training, to ensure that Bayer scientists are actually using and benefiting from the system. Indeed, Lion is partly guaranteeing Bayer's results from the new system: roughly half the total value of the deal depends on LBRI using the system on its own to find Bayer new markers and targets—500 in all.

Not Enough Bioinformaticians

Software aside, the real key to the deal is people. Bayer needed bioinformaticians, but had little chance of finding anywhere near enough. In Europe, most have been sucked up by start-ups (there are more than 20 in Europe, says Kremoser), with the lion's share going to Lion, which now has about 80 employees. Bayer now gets 20-25 Lion employees focused on Bayer's bioinformatics challenges. And at the end of the deal's five-year term, Bayer has an option to take over the group.

Lion executives say the company has another few deals teed up, but that the next several ones likely won't be as lucrative as the Bayer arrangement largely because of the tight personnel market. "The next Bayer-like deal won't be in Boston or Heidelberg," says von Bohlen, "because we've exhausted the supply [of bioinformaticians] in those areas." Indeed, says Kremoser, Lion can't easily duplicate the Bayer deal: "We can't afford to put 20-25 people exclusively on one company."

Instead, he says, Lion wants to follow the model of SAP AG, the German software star whose complex programs run finance, human resource, manufacturing and finance systems for many large companies. But SAP's multi-billion dollars of software sales generate only a fraction of total SAP-related revenues: the majority of SAP-related spending goes to implementation, customization and training contracts from major IT consulting companies like PriceWaterhouse Coopers, Ernst & Young, and Anderson Consulting. It's not enough to buy SAP's notoriously complex software; companies have a huge problem of getting the software systems working and used in the appropriate manner.

Lion's goal is to create a bio-SAP in every drug company, likewise leveraging IT consulting firms to adapt its tools to the individual companies' unique working processes and to train the clients' researchers. "We can't hire enough people to do that," says Kremoser, "so we'll go back to IT consultants" who can do the basic IT work, like building an intranet communications system, without having to learn the complex biology that Lion will focus on.

But even so, it's not clear that Lion will be able to sell enough such deals, particularly since, as von Bohlen acknowledges, middle-level research management, whose detailed understanding of the bioinformatics challenge is crucial for pushing large deals ahead, are often unwilling to share the data they have with other parts of the organization—one of the crucial benefits of bioinformatics. "It's a kingdom syndrome," he says.

Turf wars aside, such managers are interested in new sources of data from companies like Celera or Incyte Pharmaceuticals Inc. Lion therefore hopes to convince these companies to wrap their data in envelopes made of Lion bioinformatics software—in effect delivering both tools and content. With Incyte as an investor, NetGenics is a bit further up the curve in this regard: NetGenics recently completed a project it had been working on since 1997 to allow researchers to use Synergy "to retrieve, analyze, and annotate Incyte's LifeSeq sequences without altering the native LifeSeq data," according to the company.

But it's not clear that any database company is particularly comfortable with the idea of tying itself exclusively to one bioinformatics provider—particularly since, like Celera, they may be interested in providing the bioinformatics themselves. Celera is an SRScustomer, von Bohlen says, and may decide not to go much further down the road of its own overlapping bioinformatics efforts.

The Skeptics

Lion is by no means alone in its goal of becoming the bioinformatics integrator, standardizing software for the research community. NetGenics CEO Manuel Glynias likes the Microsoft analogy: "We'll build a framework that allows for ‘extensibility'—from genetics to other kinds of discovery data, all on the same operating system"—called Synergy. As soon as ten of the top 40 drug firms accept NetGenics' solutions, says Glynias, the company will have de factocreated a software standard for the industry and will thus much more easily sign up new customers. Once it has them, NetGenics will supply those clients with software and consulting, more or less like Lion.

Getting customers to buy into its system will mean NetGenics needs to supply not merely an operating system, but applications. Microsoft, Glynias argues, needed both applications and an operating system to create its dominant position: it "created a system for programs which could talk to each other and then it created enough market share with programs like Exceland Word that other companies built software for them." Thus NetGenics is also looking to create some industry-standard applications: once they've been accepted, other software developers will begin writing additional applications for their operating system.

So far, it's used its Synergyplatform as the basis for several applications, including two it's co-created with customers, one with American Home Products Corp. 's Wyeth-Ayerst Research , and one with Pfizer Inc. With Wyeth-Ayerst, NetGenics developed a tool for managing gene expression data from chips; with Pfizer it created a program for sharing data across the organization on particular gene families, incorporating five public databases and nine internal Pfizer databases.

It's also signed up as a subcontractor to IBM to write applications for and sell the computer giant's Discovery Linkdatabase system which can query multiple databases, written in multiple formats, as if they were a single database. On a single search, says IBM's McCurdy, a researcher would be able to pull from sequence, protein, and clinical databases, and from Medline.

In short, like Lion, NetGenics plans to create its integrated offering with a set of alliances. But even with alliances, it is by no means clear to observers that any one company, particularly a start-up, can create one-stop shopping. "There isno one-stop solution," says Michael Fogliano, PhD, manager of molecular sciences at Pfizer—either in terms of technology or business model. "The level of innovation must be exceptional and in all areas: infrastructure, analysis tools, and everywhere in between."

The requirements and costs of innovation are particularly high because of the sharp differences in technology focus among companies. All large businesses require finance, operations, manufacturing, and human resource management tools; but drug firms vary widely in their bioinformatics specializations, notes Robin DeMent, the director of bioinformatics for Glaxo Wellcome Inc. , the US unit of Glaxo Wellcome PLC. "SB is focused on genomics and genomic analysis; GW on genetics and genetics analysis; and Merck on gene therapy and other efforts," she says. "What works for one won't work for another." She adds: "It's going to take [a bioinformatics company] two-three times as long to get a handle on us and on SB since we're both so different and have such different sets of tools."

Even within a single company, the constant requirement for new and innovative software can make the selling of new applications extremely difficult. "Last year one company brought out a tool for broad-scale tracking of PCR sequencing and screening," says DeMent. "But we want to find a non-PCR system. And that will happen. Lots of companies are working on it. And so we haven't spent a lot of time writing PCR tools," and it probably isn't worth it for bioinformatics companies to do it either for the long term.

Nor does DeMent think it likely that bioinformatics start-ups will be able to solve their problems by relying on IT consultants à la SAP. SAP consulting is a cookie-cutter business compared to bioinformatics consulting, she argues. "The priorities [of SAP customers] are stability and manageability because all of these operations are rigorously audited. But in bioinformatics the priority is flexibility. By the time you get software installed, the science has moved on. If a bioinformatics company says: ‘we're going to give you this infrastructure and then Anderson comes in [to do the same thing it might do with SAP], as soon as Anderson learns it, they'll have to learn it again. Forget stability—go for flexibility. The whole function of SAP is opposite to the function of bioinformatics." Indeed, while Lion says it has had discussions with several of the large IT consultants, none has yet publicly signed on, committing to train its employees to work with the new system.

But ultimately the difficulty large companies face in assigning start-ups the tasks of building and integrating enterprise-wide solutions is simple trust. Few of the new bioinformatics companies have been around for very long; even fewer are profitable. Moreover, notes DeMent, drug companies "have been bitten hard" by start-ups looking to sell expensive research solutions—noting the underperforming rational drug design technologies of the late 1980s. More difficult still: "We're not going to look hard at companies who say ‘We're going to sell you all this—and we're going to control it, too."

IBM seems to understand such obstacles. It won't move upstream into the analysis of data, which McCurdy calls "the second half of the solution." Instead, it will partner with bioinformatics specialists like NetGenics—but it will by no means tie itself to one player, particularly since no one knows who the bioinformatics winner will be, or if there will be one. Regarding bioinformatics partnerships, he says, "We're from Switzerland."

IBM believes it can attract a multitude of partners because its database technology will free bioinformatics companies from writing the database end of their applications—the actual bringing together of disparate information—allowing them to focus on presenting the information itself and doing something with it. "We've created capabilities at the infrastructure levels," says McCurdy, where bioinformatics companies don't have the skills. "But we're not going to go out and hire 100 bioinformaticians. We believe in doing what we're great at." (Even so, IBM is, to at least some extent, competing with bioinformatics pure-plays. Lion's SRS, for example, is a direct competitor of IBM's Discovery Link, and is now in use by far more clients.)

Thus, while IBM likes the bioinformatics market, it doesn't like the investment necessary to bring customers the complete bioinformatics solution. IBM can make money in bioinformatics because Discovery Linkis based on technology developed for other information-intensive industries. Moreover, Discovery Link opens the way for IBM to sell other products and services. Once a company adopts DiscoveryLink, McCurdy argues, its executives might figure "they could change the way they do R&D effectiveness. Well, we've got IBM Health Care Consulting. Or it might change the way they do implementation—maybe their WAN [wide area network] won't support what they need. We've got IBM Global Services."

The Mid-Sized Market

If Robin DeMent is skeptical about one-stop outsourcing solutions for Big Pharma, she does think that it "may make sense for smaller companies who can't afford their own infrastructure."

Exactly, says Keith Elliston, PhD, the CEO of start-up Viaken, which is betting that smaller companies will look to outsourced bioinformatics solutions because they have no other realistic financial choice. "We'll bring them the capabilities they need to compete with Merck and Pfizer," says Elliston, the former CEO of Gene Logic Inc. and, before that, a bioinformatician at Merck.

Viaken wants to host almost the entire computer architecture of a full-scale bioinformatics effort, leasing space and equipment at Exodus Communications. It won't develop software, leaving that to companies like Lion, but it will host their applications. It's signed deals with Synomics Ltd. , which provides integration architecture for building user interfaces for databases, and with Oxford Molecular Group PLC 's Genetics Computer Group Inc. unit, whose SeqStore relational database allows its clients to manage genome data. They're also developing a module to manage SNP information. "GCG wants to focus its efforts selling to the top 20; we're focusing on customers they otherwise couldn't get," says Elliston.

Viaken, like other bioinformatics applications service providers (ASPs), harbors ambitions for distributing proprietary data. In the meantime, however, it aims to copy public databases, downloading them onto its own servers and then providing its clients with a completely secure portal. Drug companies keep researchers from using these public databases, he says, because "anyone can find out what you're looking at. We can let you see all the GenBankdata securely."

By the same token, Viaken could host corporate data as well—a screening database, for example, in which each compound is identified by a chemical record number but the registry for which is kept at the client's site. Without the registry, the database is unusable, and therefore completely secure.

If Elliston sees his biggest initial market among smaller firms, he argues that big companies, too, will eventually come around to his outsourcing vision. "When big drug companies look to migrating to a new system, we'll say, ‘You focus on your legacy system. We'll build the new one at Viaken and migrate it over.'"

Moreover, he contends, Viaken can play a role in easing the pain of post-merger consolidation. By locating key databases at Viaken, he says, merged companies can assure themselves that all the relevant scientists can get access to them, without having to jockey for data access through the protected turf of a jealous research executive from the wrong company.

Fanfare for the Common Life Scientist

DoubleTwist also has aspirations to become a full-service ASP. Its COO Rob Williamson describes the necessity for building up infrastructure to host the bioinformatics capabilities its Web clients will require. And the company has signed deals with Sun Microsystems and Paracel Inc. for the equipment it needs.

But the company's real innovation is its common-man approach to bioinformatics, one it is pursuing in direct competition with at least two other companies, eBioinformatics, which stems from an Australian operation; and Compugen with its LabOnWeb.com portal. "Our original model served only bioinformaticists," says Williamson, "not the real customer, which is the life science researcher." Indeed, the real customer, he says, "didn't know the power of our technologies."

It was easy to overlook him. One of the young industry's biggest problems, thinks Glaxo's DeMent, was that it focused on bioinformaticians, "not the researchers at the bench. This bled some egotistical views into the new biotechs by those who only knew the computer-side of the problem." The result: bioinformaticians talking only to bioinformaticians.

To familiarize the life scientist with DoubleTwist—and to fulfill its new mission of answering questions, not merely providing bioinformatics tools—the company is launching a portal, which Williamson hopes will become the "happy home for the life scientist." The home has been sparsely furnished so far—though it's hoping to attract Web residents with stories from an industry newsletter (Bioinformatics Publishing), stock quotes, headlines via PR Newswire, and a listing of jobs. It's also teamed up with a start-up called Oncology.com (funded by Michael Milken, Andy Grove, CMGI Inc., and ProQuest Investments LP)—though the exact fit between what is essentially a consumer site and DoubleTwist's scientist-targets isn't immediately clear.

In terms of its own core bioinformatics business, DoubleTwist is essentially offering pretty much anything it provides to anyone who wants to buy—in single-person, multi-person, or enterprise wide subscriptions to one or more tools and databases. "DoubleTwist is like a car company," says Williamson, "but we don't just sell the car. We'll sell the engine, the steering wheel, the wheels—whatever's appropriate."

And it's doing its best to aggregate as many services and tools as it can. It sells information from its own curated and annotated versions of public data, from AlphaGene Inc. , Myriad Genetics Inc. , Highwire Press and, most recently, a bioinformatics start-up called EraGen Biosciences, which sells a database of genetic sequences organized by evolutionary families.

The knock on DoubleTwist and its Web-based competitors is the financial model. "Yahoo can be successful," argues Keith Elliston, "because it's got millions of customers. But the industrial life sciences market is less than 150,000 people. If the average price you're getting is several hundred dollars per person"—and given the consumer market they're aiming for, DoubleTwist, eBio, and LabOnWeb won't be able to charge much more, he contends—"you'll need to get all 150,000 to make enough money. You need premium services." Agrees Lion's Claus Kremoser: "There's just not enough revenues in that kind of software licensing." Leveraging into E-Commerce

But the Web plays aren't simply pursuing a software licensing business, though they contend they canmake money there. EBio is pursuing a kind of per-click pricing model: a $480 subscription will buy a user about 1,000 "eBioinformatics Units," a kind of on-line currency which is spent each time a user employs one of eBio's best-of-breed bioinformatics tools (the company is aiming to aggregate about 150 onto its site—including Lion's SRS). With just 6,000 users, eBio CEO Howard Goldstein figures the company will turn cash positive, a milestone he estimates his company will reach in June 2000.

DoubleTwist is offering users free access to a fairly limited bioinformatics capability. It will charge more for multiple user access and for the system's more powerful capabilities, what Williamson calls "genomic functionality." Some companies, contends Williamson, will buy one or more DoubleTwist tools as enterprise solutions. Other users will buy individual subscriptions. "We'll have a mix of $500 customers and $100,000 customers," he says.

But DoubleTwist wants to do more than passively sell these customers software tools and data. It wants to alert users to the availability of purchasable information and products within the context of their search. Says Williamson: "DoubleTwist might tell a user: ‘We've found something in X's site that's interesting to you. Would you like to buy it?'" For example, if a scientist's DoubleTwist search leads him to a particular gene, DoubleTwist can tell him where he can buy a clone so the experiment can go on. In return, DoubleTwist could take a distribution fee orsell the user on a higher-value subscription that would net him the information he's looking for. As a first step in this e-commerce strategy, DoubleTwist has teamed up with SciQuest and Invitrogen Corp. to provide some of the reagents for this context sensitive e-commerce.

EBioinformatics is on the same wavelength. Using what it calls SmartLinks, eBio is also hooking a researcher to the logical product his experiment leads him to buy, linking perhaps to a large e-commerce catalog house or one of a myriad of much smaller vendors of high-value reagents or tools—a specific clone, enzyme or microarray, for example. EBio would get a percentage of the transaction. Goldstein says that almost half of eBio's traffic results in aSmartLinks purchase, the reason he expects vendors to eagerly sign up. Blurring the Line

While eBio sees e-commerce as the key to its Web strategy, DoubleTwist is more reticent. "We're not focusing on the e-commerce site as our business model," says Williamson, "but with finding researchers the answers they need."

Whatever the ultimate importance of e-commerce to the Web strategies of the bioinformatics plays, the boundary lines that separate this life-science segment from others are blurring. The Web tends to compromise the pricing of biological information. Too much public data is available for free; and academic researchers and start-ups like DoubleTwist are improving its quality. Likewise, the availability of shareware and the highly specific needs of bioinformaticians mitigate against the ability to charge high prices for software or, alternatively, to find large populations of users willing to pay moderate prices.

Thus the more Web-based players look to leverage their bioinformatics technologies into other areas in order to increase their revenue potential—infrastructure hosting, for example, or e-commerce—the more they compete with, or become acquisition material for, other kinds of companies. EBio's Howard Goldstein admits his company might want to forgo an IPO; a public offering might not be justified by the size of his company's market. Instead, eBio could be better off as part of a larger e-commerce play, a Chemdex or SciQuest, or even an analytical instrument or reagent supplier like PE Biosystems Group . After all, Chemdex, on the medical device side of its business, is wrapping its e-commerce distribution play, Promedix.com , with information supplied by its latest acquisition, SpecialtyMD.com .

Likewise, database players like Celera, which sees bioinformatics as a critical way to tie customers into long-term contracts, might want to buy a bioinformatics player. And given the enormous valuations the stock market is putting on genomics and e-commerce players, acquisitions of the still mostly private bioinformatics players are eminently possible.

That such cross-segment consolidation which, only a few years ago, would have seemed absurd, now seems the most likely end for most of the bioinformatics players simply underlines the basic business problem of bioinformatics: generating an adequate reward for the investment. In an industry like bioinformatics, where the cost of developing a broad set of world-class offerings is high (in part because no software is ever finished in the constantly changing world of drug research) but the value uncertain—no one has ever developed a drug using a bioinformatics program—companies need to closely tie themselves to products.

E-commerce is thus a logical end-point for bioinformatics. So is pharmaceutical research itself. It's no coincidence that Lion—which is guaranteeing that its software will uncover 500 targets and markers for Bayer to research—is now setting up its own discovery operation with some German universities, trading software for rights to nuclear receptor targets. Until start-ups can routinely convince drug companies that bioinformatics software deserves the same kinds of prices paid for the databases they analyze, bioinformatics companies will need to hedge their bets with more obviously remunerative businesses.