IBM Dives into the Life Sciences

IBM thinks that the life sciences industry, where it has had little presence to date, is now big enough and ripe enough to be worth pursuing. It aims to get in at the start of the value chain via bioinformatics. The computing giant has bigger plans than smaller players that moved earlier to focus on bioinformatics-and it has the means to see them through. Pioneers got stuck doing fee-for-service work, but IBM can sell Big Pharmas hardware, software and most importantly, all sorts of services. IBM's debut product in life sciences is DiscoveryLink, which lets data from disparate sources be queried as though they were all in one, giant database. It's also meant to broadly support software created by specialized applications developers. Installations will have to be custom jobs-ideally, part of bigger IT contracts. Other computing concerns also perceive opportunity in life sciences; indeed it's clear a battle is brewing between big hardware suppliers. But they're cooperating to a degree: calling for data standards, so they can compete on products not technology. IBM has credibility from other sectors, but it's uncertain how applicable it expertise will be in life sciences, even if it can find and train enough people. Also unclear: whether drugmakers actually want and will pay big money for integrated solutions. Though a newcomer to life sciences, IBM may be uniquely suited to serve the industry-not only because of its own deep and ongoing research into computational biology, but because it can offer drugmakers one-stop shopping.

IBM is going where bioinformatics companies have never been before: into the realm of large-scale service, to help life science firms make the most of their data.

by Deborah Erickson

• IBM thinks that the life sciences industry is now big enough and ripe enough to be worth pursuing. It aims to get in at the start of the value chain via bioinformatics.

• The computing giant has bigger plans than smaller players that moved earlier to focus on bioinformatics—and it has the means to see them through. Pioneers got stuck doing fee-for-service work, but IBM can sell Big Pharmas hardware, software and services.

• IBM's debut product in life sciences is DiscoveryLink, which lets data from disparate sources be queried as one. Installations will have to be custom jobs.

• A battle is brewing between big hardware suppliers. But they're cooperating to a degree: calling for data standards to even the playing field.

• IBM has credibility from other sectors, but it's uncertain how applicable its expertise will be in life sciences, and whether drugmakers actually will pay big money for integrated solutions.

• Though a newcomer to life sciences, IBM may be uniquely suited to serve the industry—not only because of its ongoing research into computational biology, but because it can offer one-stop shopping.

In 1996, in a move that surprised many people in the pharmaceutical industry, International Business Machines Corp.(IBM) acquired The Wilkerson Group (TWG), one of the leading consulting firms in the sector, for a few tens of millions in stock. The Wilkerson acquisition seems to have paid off: thanks to leads from TWG, IBM signed consulting deals worth several hundreds of millions of dollars each with Bristol-Myers Squibb Co. and Novartis AG , before closing down the operation—most of whose assignments were, by IBM's standards, too small to bother with.

Now IBM is pursuing a similar strategy to get into the pharmaceutical research and development world—it's using a fairly low-cost entry point in bioinformatics, to wedge its way into large drug companies and the biggest biotechs. IBM management has noticed that the work of these companies is becoming ever more data intensive, and has decided that the life sciences industry—where it has had little presence to date—is now big enough and ripe enough to be worth pursuing. The computing giant wants to leverage the capacities that make it a leading supplier of information technology services in the life sciences sector. IBM would like to sell drugmakers integrated solutions that could entail up to hundreds of millions of dollars worth of software, storage capacity, hardware and services.

IBM is diving into the life sciences via bioinformatics—a term that now stretches to fit computing-based approaches to processing all sorts of data needed and generated by life science researchers, including the wealth of data in sources that they have tapped only haphazardly so far, such as academic literature and lab notebooks. Public data sources, such as GenBankand the Protein Data Bank, that were rapidly filling up with data a few years ago are now brimming. Researchers want to make the most of all that, as well as the data spilling out of lab instruments such as mass spectrometers and tools such as gene chips.

Theoretically, the market potential for bioinformatics is growing, along with the assemblages of data that drug companies consider integral to their research. Human eyes and brains simply can't process all the data available today; we need information technology to make sense of it all. But that's just what venture capitalists were saying in 1996, when they began setting up bioinformatics companies like Molecular Applications Group (MAG), Pangea and NetGenics Inc. These early bioinformatics firms have not fared well.

Now the marketplace has gotten even tougher. Drugmakers are far less willing to pay big dollars for technology solutions than they were just a few years ago. They want the compounds themselves, and they're only paying for those on a milestone basis. Executives in the bioinformatics business have yet to distinguish the value in their offerings from other discovery platforms like combinatorial chemistry or knockout mice.

Bioinformatics companies that set out to serve the industry with software-centric models found that they'd gone off on too narrow a path. The algorithms and computer programs they offered gathered and analyzed some specific sorts of data. They solved problems that were undoubtedly important to some researchers, but were seen as just a piece of the drug development process. That's why the model was and remains limited. Nothing any bioinformatics firm has yet come up with has been proven to solve the great mystery of pharmaceutical research productivity. Until someone can credibly claim to be able to do that, software-centric bioinformatics firms stand little chance of becoming high-margin businesses.

Enter, IBM. It has bigger, broader and deeper plans than smaller players that moved earlier to focus on bioinformatics—and it has the means to see them through. IBM thinks that lack of data integration is the bottleneck keeping drugmakers from making better use of the data they've gathered. That is the premise guiding its move into the life science industry by way of DiscoveryLink, a form of software called "middleware" because, like Windows, it sits between an operating system and raw data sources, and supports specific applications. DiscoveryLinkputs a wrapper around databases so that the contents can be translated into a text-based programming language called XML, regardless of what format they're actually stored in. With DiscoveryLink, data from disparate sources can be queried as though they were all in one giant database. It's supposed to save time and reveal new insights.

IBM doesn't plan to sell application-oriented software—the kind of tedious, difficult-to-make stuff that the start-ups held themselves to doing. If the industry evolves as IBM would like, there will be many such programs written by specialists and running on many computers in systems it has helped devise. Even the middleware IBM has created doesn't have to be its main value-generator, because the firm also provides a whole host of services, hardware, and data storage and management options that will drive up the total price tag of its offerings to any one client.

In effect, IBM is saying to drugmakers, "We've got a lot of experience devising IT solutions for other industries and we think you people in the life sciences need to integrate your data in order to get more value from it. We've made DiscoveryLinkto help do that, and we are working on other things to help your drug discovery efforts become more efficient. We've allied with some talented applications providers and are recruiting more. In the meantime, we can build integrated solutions for big players—and they'll have to be custom—but you can trust us to support you all the way." It's a compelling argument, and one that IBM is uniquely qualified to make. If it can find and hire enough people capable of doing the customization work, the firm is economically set up to support big endeavors, because it can supplement its deliverables with high-margin equipment sales and contracts for additional services such as data storage and overall data management. IBM is also considering new services for drugmakers, including those that aren't so big: it might, for instance, lease time on high-performance computers.

Small bioinformatics players like Pangea and NetGenics had nothing so comprehensive to offer—at least, no one in the pharmaceutical industry believed they did. But people are willing to believe that IBM has a ton of capabilities that may allow it to prosper through bioinformatics, and serve the life sciences industry in a very welcome way, even if it is a newcomer. Already it's gotten a few customers, announcing deals in June with Schering AG and Aventis SA . No financial terms were disclosed, but DiscoveryLink will be a key part of the solutions that IBM will implement, with help from specialist bioinformatics firms NetGenics and Spotfire Inc. , respectively.

IBM's credibility is solid, but it must still deal with some uncertainties: How applicable will the expertise it has garnered in other sectors actually prove to be in support of life science research? More importantly, how many Big Pharmas and large biotechs actually want and are willing to pay big money for the integrated solutions IBM is offering to custom-craft? Schering and Aventis have hired IBM, but other pharmaceutical companies could decide that data integration isn't a cost-effective investment now. They might prefer to charge their internal IT departments with handling that task, and see what happens. Drugmakers could also work with another supplier like Sun Microsystems Inc. , which relies on Accenture and KPMG as well as smaller boutiques to help customers build appropriate IT systems.

How widely IBM can wedge its way into life sciences will depend on drugmakers' willingness to accept the computing firm's first contention that data interoperability is vital; as well as a second, that IBM should be awarded the job of building integrated solutions. It won't be easy for IBM to persuade companies and researchers—who've always believed in the primacy of proprietary information—that data should be shared across networks. But if anyone can make that case, it's IBM, because it has everything it takes to be a one-stop shop.

Why Early Players Fared Poorly

IBM is coming into the bioinformatics market from a different angle than the early players, because it carries a much broader set of capabilities. Despite the differences in size and upside, the giant computing firm needs to understand why the companies on whose shoulders it will stand fared poorly.

When the first commercial bioinformatics companies launched in 1996, it was already apparent that the quantities of data piling up in labs were beginning to overwhelm scientists' abilities to extract information of value from them. (See "Data, Data Everywhere,"IN VIVO, July 1997 [A#1997800152.) The early firms were founded to help drugmakers deal with the data deluge…basically, however they could. Venture capitalists decided it would be okay for the small firms to begin by offering very specialized software: start-ups would "get close to their customers," gain understanding of their needs and evolve accordingly. Never mind that potential customers all seemed to want something different; the VCs decided they had to be in it, to win it.

The catch-as-catch-can strategies of the early movers might have worked just fine, if the market had been ready, asserts Manuel Glynias, CEO of NetGenics. But with hindsight it seems the market may not have been ready, for several reasons. A few years ago, data integration problems were looming but not necessarily on the doorsteps of most companies. Some Big Pharmas decided they could wait to deal with the issue, while others like SmithKline decided they'd be better off spending to hire their own bioinformaticians—a rare breed of programmer at the time. Indeed, five years ago many industry players were grumbling that SmithKline had hoovered up all the specialists.

The start-ups' pricing may have been a factor, too—particularly since bioinformatics firms were coming in on the heels of other discovery platforms like combinatorial chemistry. Big Pharma customers, rueful about spending so much on technologies that didn't deliver the gains they'd hoped, became more cautious. Looking hard at what they'd be buying for what price, they saw that start-ups didn't have anything ready-made. Nor did they have anything startlingly new enough to warrant the high prices they were asking. Researchers had grown accustomed to getting pretty good software for free off the Internet. Indeed, Glynias suspects that the SRSsoftware package marketed by Lion Bioscience AG spread out so fast because it used to be free. The German company bought the program from the European Molecular Biology Laboratory (EMBL) [See Deal], and started charging for it only after many scientists had become familiar with it.

A third factor influencing market readiness related to the perceived viability of start-ups. Large firms couldn't bring themselves to trust tiny start-ups to handle the responsibility of developing enterprise-wide software solutions. Drugmakers knew all the work they'd commission would be custom, and they were fearful that small companies might go out of business, leaving them high and dry. Big firms also did not trust that small companies would be able to grow large enough and fast enough to provide training and support for their specialized software.

Unable to close sales but needing to generate cash flow, the start-ups began working piecemeal to show prospective clients that their solutions were truly valuable—different than the cheap stuff they'd downloaded from the Internet. Software couldn't command high prices unless it came with a high level of customization, training and hand-holding. The young bioinformatics business started turning into one based on a fee-for-service model.

IBM isn't going to travel the narrow path that the early bioinformatics players did. It is moving into the business via DiscoveryLink, a version of its DB2relational database—sold across industries to some 5,000 customers—that has been specially adapted for the life sciences industry. Popular public data repositories like GenBank and Swiss-Prot and Medline databases, are prewrapped, because IBM knows researchers depend on them a lot. Likewise, DiscoveryLink can also wrap any type of commercial relational database, such as Oracle's, because many life sciences firms store their in-house data that way. The wrappers can be extended to any data source that individual clients want to access.

In some senses, DiscoveryLinkis the next best thing to an operating system like Microsoft. Designed to sit above an operating system but below applications, it can spread broad and low. "DiscoveryLink may or may not drive profits at IBM, but it allows the company to establish a presence in life sciences," observes Colin Freund, VP of business development at DoubleTwist Inc. , formerly Pangea. (DoubleTwist is now focused on developing data and tools that researchers can use to do target discovery and validation.)

IBM has been working on DiscoveryLinkin one form or another, and under different names, since the mid-90s. In fact, some bioinformatics experts say this is IBM's third shot at launching the product. Industry observers note that IBM had a substantial presence at BIO this year, talking up the new and improved version of DiscoveryLink. Previous efforts to get DiscoveryLink taken up in the field were half-hearted and faded away, say bioinformaticians who have been in the field for many years. "This is the first time the company is really ‘product-izing' it, and putting marketing muscle behind it," one source says.

Like its predecessors in bioinformatics, IBM is also going to start out custom-crafting solutions for clients, albeit at a far larger scale. It has little choice, given how bioinformatics has evolved within the pharmaceutical industry as a hodgepodge of purpose-built programs and data sources, different for every drugmaker. Indeed, the heterogeneity of databases—different content, and different ways of looking at it—demands customization, Manuel Glynias points out.

What's Different Now

IBM is coming into the life sciences research market at a time when drug companies are as skeptical as they've ever been. Researchers have gotten plenty of data; they want to use it well and they don't want to waste it. Information-based companies that want to provide high-value solutions are getting the message, realizing that they need to move to either a put-technology-together-so-it-makes-sense phase, or actually provide a compound. The real problem for drugmakers is not getting information; it's getting new drugs.

Information is no longer selling for high prices. That's a big problem for companies that have only information to sell, or worse, technology or software meant to generate information, but not yet proven to produce something of genuine, actionable value. IBM has to confront this market reality, but it's got ways around it. Like Wal-Mart, it can provide a certain thing (i.e. software) relatively inexpensively because it also sells other things drug companies need. The one-stop shop model is well suited to tough market conditions because the margins on different offerings even out, making any one thing just part of the mix, not the live-or-die, must-buy item.

Customer demands may have gotten tougher, but certain aspects of the bioinformatics business have gotten easier in recent years—enough so that IBM thinks it can surmount the obstacles that have thwarted other bioinformatics players.

IBM is confident that large drugmakers will want the kinds of large-scale solutions it hopes to provide, because data that's been piling up for years is now reallypiling up. Data streams are becoming ever more unmanageable, and the need to get the situation in hand is more obvious. The more data people see, the more they realize that they can't simply invest in data-generating technologies unless they're simultaneously learning to use data-crunching technology. That realization has led to a dramatic slowdown in the number of deals for data-generating technology. But companies still have to deal with the data they've got, and that which they'll accumulate via the deals and sources they already have on tap.

One key factor that may help IBM solve clients' data integration problems now is the way programming languages have evolved over the past few years. When bioinformatics was new, the experts pushing the field forward were working to develop object-oriented programming approaches. They were thinking of the mechanics of data exchange, thinking about data sets as shapes that resided in a computer and how they related to one another. But it turned out that CORBA (Common Object Request Broker Architecture) was just too complicated to become pervasive. "You had to be a rocket scientist to use it," one source says.

Key members of the IT groups growing up within Big Pharmas outright refused to work with the object-oriented language, not only because it was extremely complex but "because it didn't work well through the Internet." That's putting it mildly. In fact, CORBA had been deliberately designed to operate through a dedicated port into a computer—precisely the channel that IT staff in big companies shut down, as the use of computers and the Internet continued spreading throughout US industry. Companies that bit hard on CORBA, like NetGenics, had to reorient themselves and build new systems.

These days, people writing bioinformatics programs are tending to work with a text-based way of representing data, called XML. It's not a standard language yet—there are at least 13 versions in use within the drug industry now, and people are starting to stray away even from those, using different types of vocabularies. But industry observers say XML has much more potential to become a standard than CORBA ever did, and point to some beginning convergence: at the end of June both EMBL and Millennium Pharmaceuticals Inc. announced they'd agreed to write their future programs in XML.

Desperately Seeking Standards

IBM would like all software companies and pharmaceutical companies to agree to make any data they generate or process exportable from its source in a common format—namely, XML. That's the key thing IBM will push for as a member of the Interoperable Informatics Infrastructure Consortium (I3C) formed at the end of June, at BIO, the giant life sciences conference held in San Diego. Members of the new consortium include Accenture, LabBook Inc. , the Whitehead Institute for Biomedical Research of the Massachusetts Institute of Technology (MIT), TimeLogic Inc., Millennium, Oracle, and the National Cancer Institute .

IBM wants standards for the same reasons that the other information-consortium members, including organizer Sun Microsystems, want them. If software developers agree on some basic standards for bioinformatics data, it will level the playing field so that everyone in the field can compete solely on a product and service level, instead of on a technology level.

But any wide-reaching agreement on standards is far off. At best, standards may be implemented bit by bit, because the I3C—which originally set out with very broad objectives—soon decided to bite off more manageable chunks of the problem. The group aspires to have standards for analyzing gene expression data within 15 months. This modest start reflects past industry experience with standard-setting bodies: there's been discussion over years and years, but nothing has ever gelled.

What's to spur companies to pull together on standards now, when they never have before? Caroline Kovac, VP of IBM Life Sciences Solutions (a business unit IBM formed in August 2000), believes drugmakers will agree to embrace standards, because "the point of pain is increasingly severe. Pharma companies are really caught on that point." Kovac previously held several top management positions within IBM Research, including VP of technical strategy and division operations and head of the firm's computational biology efforts. Kovac says, "Pharma companies are saying, ‘We cannot continue to build our efforts in discovery at the pace we have to, with new information sources coming online every day, unless we have a standard architecture.' Pharma companies are saying to applications providers, ‘We want you, but you've got to interface.'"

But the fact is that no pharmaceutical companies have declared themselves members of the new I3C consortium, and their absence is noteworthy. Until drugmakers start publicly saying that they will not buy bioinformatic products that do not conform to a particular standard, there will be no motivation for commercial programmers anywhere to conform. IBM acknowledges as much: "It's up to us to create a value proposition that will encourage application providers to conform to a standard," declares Jeff Augen, PhD, director of business strategy at IBM Life Sciences.

To show its commitment to the I3C, IBM donated to the consortium 14 detailed descriptions of the way data flows when someone is retrieving sequence data across the Internet, or distributing a piece of content from an internal database to a customer. "These descriptions are normally intellectual property that people don't give up—it's hundreds and hundreds of pages of analysis," Augen points out. Handing the work over to the consortium clearly wasn't an altruistic act on IBM's part. Its descriptions detail IBM's way of thinking about processes that programmers must have figured out in order to write software to influence the process.

Sharing the descriptions was also a move to give programmers, academics and bioinformaticians-in-training a guided start at creating software that will plug into DiscoveryLinkand help carry it into the life sciences industry. It's a symbiotic relationship that IBM envisions forming with applications developers. Specialists that bundle DiscoveryLink into their products get some credibility-by-association from doing so. Caroline Kovac says, "DiscoveryLink has two sets of customers, and one of them is a channel to the other."

IBM is straightforward with potential collaborators about the kind of cooperation it expects. "If people say they don't want to embrace XML because they think XML storage takes too much space, we say the idea is not to store in XML but to translate outbound into XML. If you're not compatible with that, you're not compatible with us," Augen explains. He says, "That's the message, and it's the same one that other consortium members like Sun will convey. If you want to use your own standards for storing data, then you won't be able to expose the right interfaces to us and you won't be able to work with us."

IBM continues scouting for and allying with applications developers that it thinks are now or could become best-of-breed providers. (See Exhibit 1)Executives of the small firms say they're happy to be allied with IBM. Structural Bioinformatics Inc. , building a system to do 3-D protein modeling at atomic-resolution, is one of those IBM has picked for a winner—good enough to become its first equity investment in November 2000 [See Deal]. CEO Ed Maggio says he thinks IBM's Life Sciences group has understood, "in a visionary way", the issues that will be important in processing and manipulating large scales of data.

Maggio says SBI wanted to work with IBM in part because data privacy is so vital to its business model. "We're developing very accurate structures, not available from any other source, and selling them for a substantial price. So we want to be sure only our customers will be able to obtain the data," he explains. Likewise, he knows that the biotech and pharma companies that are SBI's customers have equally strong concern about communication between their facilities and external databases. No firms want their competitors to become aware of their research interests. "It's a valid concern—and IBM is a leader in secure transactions," he points out.

IBM has also begun working with several leading venture capital firms to help it forge connections with the right companies. About eight months ago, the company formed strategic relationships with Oxford Bioscience Partners, headquartered in Boston, Burrill & Co. of San Francisco, and Sofinnova Partners in Paris. Augen says IBM hopes to benefit from their deal flow to get "headlights" into future technologies, such as biochips. The company also hopes its partners will let it know if they spy firms that could grow much faster if given a piece of technology or other support from IBM—even if the VCs aren't investing there.

If IBM picks well, then the fact that applications truly appealing to life science researchers are built to standards that IBM supports and run on its systems should further drive acceptance of the standards and also help IBM wedge deeper into the sector. Without standards, drugmakers will have to go on custom-building connections each time they want to hook into a new data source. And even the best applications won't be able to spread as quickly. IBM clearly isn't waiting for agreement on standards. It's getting into bioinformatics now: indeed, one observer describes the huge company's fast entry into the field as "a big belly flop into the pool: ‘We're here!'"

Computing Concerns Compete

IBM isn't alone in sensing the opportunity to get into a supplier relationship with life science companies through bioinformatics. In fact, there's a war shaping up between the big hardware-based computing firms.

It was Sun that organized the I3C to champion open standards for bioinformatics. Before that it backed Web standardization, having organized the W3C, the World Wide Web Consortium, which agreed to the protocols (Java) that enable Internet browsers like Netscapeto do interactive business online, rather than simply provide a look at information.

Sun has been thinking about the role of IT in the life sciences industry for a long time, asserts Siamak Zadeh, Sun's group manager of life sciences—who believes its groundwork will pay off commercially. Sun put together the Life Science Informatics Advisory Council, a group of experts from the pharmaceutical and agricultural industries and academia in December 2000. It gave rise to I3C, having concluded that the lack of standards, or more accurately the existence of multiple ones, was keeping the bioinformatics industry from advancing commercially.

IBM is "a newcomer to life sciences," Zadeh sniffs, adding, "They are following our strategy, which is why they're talking about partners now when they never did before. They're going to learn the hard way that destructive policies—playing down the merits of other companies' approaches—always backfires." Some observers of the bioinformatics industry think Sun has done a smart thing, once again championing open standards and getting the recognition for doing so. That openness is the reason a lot of software has been written for middleware and software that runs on Sun machines—like Oracle software, which currently supports by far the largest installed base of relational databases in the life sciences industry.

Some industry insiders say that Sun "should have had the lead in this sector," since it has been serving it so long. Perhaps the company was so busy watching the struggles of small bioinformatics companies that it failed to notice the advancing shadow of mammoth IBM, which suddenly started moving fast in the past year. Sun may also have discounted IBM as a life sciences competitor because previous half-hearted attempts at developing DiscoveryLinkwere not so successful. Others say Sun's apparent slowness has been a big networking success, allowing it to make and develop a lot of contacts.

This much is clear: Sun is not going down without a fight. Compaq Computer Corp. is also a contender for share in the life sciences market. It currently sells the fastest processor in the industry, and would clearly like to do more deals like the one it did with Applera Corp. 's Celera Genomics Group —supplying Alpha processors to other drug developers. Industry observers say Compaq has "the fastest, screamingest processing capacity, but not much human capability behind it," and is "struggling a bit with its identity." Zadeh figures, "Sun and IBM will be top players in the life sciences market, which is huge, certainly big enough to share. Probably not Compaq, which is only about hardware." Still, the company can't be discounted.

IBM is already swinging at Oracle, whose relational database is unquestionably dominant in the life sciences industry. "They're clearly entrenched," Caroline Kovac says, "but they've got the wrong product." Once that recognition strikes users, Kovac believes drugmakers will begin loosening their connections to the company whose "answer to the problem of data integration has always been, ‘put it all in an Oracle database and manipulate it there.'"

The problem, she charges, is that data sets are getting too big to import into Oracle. Jeff Augen argues that it's just not feasible for pharmaceutical companies to do daily updates of GenBankinside their firewall any more, noting that the database is growing by 850,000 base pairs per hour. Even when drug companies want to make the effort to bring large quantities of data in-house, they can't always do it, he observes: "You can buy a subscription to Celera's database, but you can't get a copy of it." He reckons that other data sources will evolve similarly, compelling people to reach out to data without being able to bring it into their own systems.

IBM is betting that the industry will eventually become accustomed to tapping into data that is stored in separate sources, rather than importing it—particularly if standards make all content uniformly exportable. Some data sources already charge for access to their content on a "per-drink" basis, and IBM is prepared if the trend spreads within the life sciences industry, Kovac notes. It has developed software for other industries to facilitate similar billing and payment transactions.

It's too soon to say how the fight will be fought, let alone who will win coming battles between computing-based firms that already have a position in life sciences, and others like IBM that are just beginning to wedge their way in. The bioinformatics market is still so young and mutable, that new competitors and allies are appearing all the time.

Very recently, Cray Inc.announced that it too is interested in working with life science companies. However, it proposes a very different approach from the distributed-network systems that IBM and Sun promote. Steve Conway, Cray's VP of corporate communications, says the company only just realized the potential already built into its Cray SV1 supercomputer. Not six months ago, he says "one of our senior programmers who is adept at working with the intelligence community in the US and allied nations was watching a 60 Minutes program on genomics, and had a Eureka moment: he realized the problem types were identical."

Cray quickly phoned its contacts at the National Cancer Institute, which has always used Crays for it heavy lifting, and shared the insight. A visit and some experiments later, Conway says Cray started getting reports from NCI that problems which took seven days to run with Compaq's Alpha Processorcould be done in 130 seconds with the Cray SV1. And no wonder: an Alpha reads 69 million characters per second; a Cray SV1 reads 9 billion characters per second.

Some potential customers are enthusiastic about applying Cray's processing power to life science problems. Others are not, telling Conway that bioinformatics "just isn't done that way." He says, "the short message we're getting is, ‘If you want to be in this business, you have to do things the same way we do.' We're saying, ‘Let's talk. It could be done differently. If you're a company whose success depends on the ability to solve problems better, faster, cheaper, let's talk.'" Unlike IBM, Cray does intend to write applications. The firm has been doing it for 20 years, and is known to be strong at that.

Meeting Requirements at Aventis

As companies like Cray contemplate their own entry points to bioinformatics, IBM is honing its data interoperability pitch to large pharmaceutical firms it believes will benefit from custom-crafted solutions. It's concentrating first on drug discovery, but certainly looking farther.

Life science customers exist along a continuum that stretches from the laboratory to the market, and IBM would like to work with as many of them, at as many points on the path, as possible. The company is betting that its first product, DiscoveryLink,will help it get in at the very start of the value chain, in discovery research, where it recently signed up two Big Pharma clients.

IBM is working hard to build the trust that will help it move down the line—and the going isn't always easy. Dominique Hervé, global head of Aventis's Target Lead Candidate portal project, says that IBM's initial offering to it—a custom-crafted solution involving DiscoveryLink—was a disappointment, in fact unacceptable.

Senior IBM representatives had pitched top management at Aventis three years ago, just after the merger of Hoechst Marion Roussel and Rhone-Poulenc Rorer [See Deal], and got an immediate go-ahead to implement a solution that IBM said would allow different research sites to communicate with each other. "They said it was wonderful and ready, but it was not," Hervé says.

The fault was not all IBM's, he says: Aventis had not clearly defined its requirements. But Aventis still needed an IT solution, and Hervé decided that the best way to get help was to make a request-for-proposal to eight companies, and see what they proposed to do, at what price. After an initial survey of potential competitors, Hervé issued invitations to present to companies of various sizes, including Compaq, Tripos Inc. and a tiny French firm called EXML Media.

Some firms were ten times more expensive than IBM, others less expensive, and some were totally unprepared, seemingly interested only in how big Aventis's budget might be. One firm lobbied aggressively with a contact within the firm, thinking that person would pressure Hervé into a decision. Ultimately, the firm that fulfilled Aventis's requirements was IBM, he says, emphasizing that for a different company, a different supplier might be better.

It was hard getting IBM a second chance within Drug Innovation and Approval, (Aventis's term for R&D), after the initial disappointing experience, Hervé says. "You can imagine, after telling everyone some years ago, that IBM has proposed a very innovative solution and we are going to have a marvelous tool…and then we had to go back and say, ‘I'm sorry, it did not work'…my VP, six months after making that announcement, was not so keen on this. I had to defend why IBM was the best solution, for a whole day."

Hervé made the Request for Proposal (RFP) in November 2000, and made his selection of IBM at the end of January 2001. It took two months to finalize the deal and the second pilot ran six weeks. At the end of the pilot, Aventis decided IBM's solution was successful and fulfilled 100% of its expectations. Roll-out of the solution began in early July and should be finished on four sites at the end of August, at which time Paris, Frankfurt, Tucson and Bridgewater, NJ should all be hooked up and able to share data.

Strength of a Matrix

DiscoveryLinkis IBM's most tangible offering to the life sciences industry, but Jeff Augen points out that because IBM is a matrixed organization, the Life Sciences business unit can draw on the resources of the worldwide firm. In addition to 80 computational biologists, another 100 people work in Life Sciences under Carol Kovac. Beyond that, in each one of IBM's brands—servers, software and global services—there is an additional person or group of people that focuses on life sciences.

When people from Life Sciences went to the storage division, and defined some specific requirements they'd have for certain devices, such as band width and operating systems to manage large numbers of small files, "The storage people said, ‘Ah, we need to develop an enhancement to our file system,'" Augen recalls. "So even though we only have a couple of hundred people in Life Sciences per se, somewhere in IBM there's a team that's working on our storage issue, and other teams on other projects. We're leveraging hundreds of millions of dollars of resources."

IBM can also claim to be wrestling with some of the very same research problems as firms bent on drug discovery. It's been doing so, not because it wants to be in the same business, but to learn more about what life science researchers need. The company started working on computational biology problems in 1992, says Joe Jasinski, PhD, senior manager, Computational Biology Center. The company has, for instance, built a suite of algorithms for pattern recognition it named Tyresius, after the blind seer from Greek mythology. IBM has been pursuing functional genomics and modeling with MIT's Whitehead Institute. That collaboration concentrated on the wet biology aspects of gene expression analysis: assessing tissue taken from lymphoma patients with chips made by Affymetrix Inc. But IBM has also been working on data specific challenges of that work with Harvard University , Stanford University and MIT.

At IBM's Almaden research facility in California, where the early programming language Fortran was invented, a team of scientists works on basic problems associated with data management and integration. Its efforts to develop tools to mine biological information fed into the creation of DiscoveryLink.

IBM continues getting press for its attempts to build the world's most powerful computer, Blue Gene, expressly to figure out how proteins fold themselves into their distinctive three-dimensional shapes. The company—and many people in the pharmaceutical industry—are eager to see if the massively parallel computer, expected to be finished in 2003, can do it. Protein-folding is one of the major questions in life sciences today, in part because it has big commercial applications: any information about the shape of proteins can help drugmakers design more selective, effective drugs. IBM researchers have been working for years on algorithms to predict structure from sequence.

IBM is also working with NuTec Sciences Inc. , for whom it agreed to build the most powerful computer it has ever supplied outside of government. The collection of 1,250 IBM servers, capable of 7.5 trillion calculations per second, is valued at $70 million, but IBM reportedly accepted the contract to build it for $10 million. The companies recently began putting it to the test in the life sciences, harnessing its power to investigate how specific gene combinations influence the development of cancer and patients' responsiveness to treatment. Some clinical work began recently at the Winship Cancer Instituteof Emory University in Georgia.

If maintaining its own research activities is a way of showing the life sciences industry that IBM has in-house expertise, it may also be a sweetener for deals. Companies that enter long-term contracts with IBM may be able to get a little extra help with their problems from its researchers and Blue Gene, says Sharon Nunes, PhD, who headed IBM's computational biology effort for eight years and just a few months ago was named director of solutions development for the Life Sciences unit. "We like being able to say that we can bring things to the table that other firms can't," Nunes declares, noting that IBM wrote some algorithms for Monsanto, since it had a long-term contract with the agricultural company.

What IBM does not bring to the table, managers say, is an insistence on installing its own hardware and software. IT industry observers confirm that IBM has lived up to its promise to be agnostic on both those fronts, when putting together computer systems in other sectors. To a large degree, it is the company's corporate structure that prevents it from forcing its own hardware and software on clients. Because the business units each have to meet their own goals for revenues and profits, IBM staff working on the various components of a big contract are motivated to give clients whatever works best with their current systems. It can't afford not to install the things customers want, on the machines they want. So IBM will put Oracle databases on its hardware, just as IBM's own DB2databases are sometime put on Sun machines. IBM supplies such a wide range of hardware that Sun's Sia Zadeh suggests the company could run into trouble supporting so many varieties.

Differentiating on Services

IBM knows it faces competition in hardware from the likes of Sun, Compaq, Hewlett-Packard Co. and perhaps now Cray; in software from Oracle and Microsoft and SAP; and in consulting from big firms like Accenture and KPMG. But the company feels nobody can compete with it, when it comes to services. If Global Services were separated out of IBM, it would be the world's largest full-service global IT firm, asserts Don Cotey, worldwide services executive at IBM Life Sciences Consulting and Solutions. Last year, Global Services generated $35 billion of IBM's $85 billion in revenues.

Global Services allows IBM to brings one-stop shopping to life science companies—a proposition that Cotey says "seems to be resonating. Drug clients would prefer to work with fewer partners than more, and on a more strategic basis." The capacities of the business reside in four separate service businesses. The Business Innovation Services group houses industry consultants, systems architects, application developers, and all different types of IT specialists to help companies apply e-business practices to custom-improve their business.

"The life science customer is challenged with accelerating innovation, turning new sets of chemical, genomic, and structural data into treatments. That's right in our sweet spot," Cotey declares, adding, "We know a lot about how individuals and teams do innovation, and the role that IT plays in supporting innovation."

IBM's Integrated Technology Services provides the capabilities that life science players will need in infrastructure, security, high performance computing, and networking. "It's a line of business a mile deep," he declares. The firm's third services group is strategic outsourcing, and Cotey says it is already the number one provider of outsourcing to the global pharmaceutical industry. It's now beginning to work with some big biotechs as well.

When drug companies sign up for Integrated Technology Services, as firms including Aventis, Novartis AG, and AstraZeneca PLC have done, IBM helps the clients figure out the kinds of systems they need, trains people to maintain those systems, provides support services and manages advanced technology products, such as Web-type solutions, complex servers and supercomputers. These are billion dollar and multibillion dollar deals for IBM. Aventis, for instance, entered a 10-year agreement with IBM a year ago—separate from the deal involving DiscoveryLink, announced in June—that's worth $1.5 billion over the duration. "IBM is somewhat uniquely positioned to be able to handle that," Cotey says.

Global Learning Services devises distributed learning and training solutions of the sort that Cotey reckons a lot of research scientists are going to need, because genomics and proteomics are so new to everyone.

The groups within the services division are highly experienced, Cotey notes, but IBM also understands high-tech R&D from its own research efforts. He says the firm is to the point of having best practices that life sciences customers get excited about, when they come to visit IBM sites around the world. IBM shows visitors how it applies its own advanced IT solutions to the research process: to improve collaboration, for instance, and let communities or groups of scientists do large-scale synchronous and asynchronous research ‘round the clock'.

Cotey believes the process of life sciences research will become more and more collaborative, noting that large pharmaceutical firms already often team up with a handful of biotech companies and some academics, maybe even some government collaborators. IBM is expert at developing business partner networks and managing discovery activities in online environments, because it's done that for itself. "It's not unique to us, but it's a way of working that the pharmaceutical industry is just moving into in the last couple of years," Cotey says. Mergers have driven the practice to a certain degree, but he sees the trend towards working in groups as a more general one.

The examples Cotey cites show that IBM is clearly geared to provide services at scale. The company has, for instance, helped NASA set up IT structures that allow it to bring together hundreds of companies, including big ones like Boeing and McDonnell Douglas, to do projects like building the space shuttle. But large pharmaceutical firms don't currently work in huge public networks. Although drugmakers might like to have better connections between their research sites so scientists might quickly learn whether anyone else within the firm has experience with a particular compound or cell line, those desires have to be balanced with concerns about keeping proprietary data proprietary.

Big Pharmas work at huge scale, but much discovery research gets done at small organizations. What, if anything, will IBM do to help these innovators, who also cherish privacy, to become more efficient through IT? Small firms seldom appear on the Global Services radar screen, Cotey admits, but he figures that many biotechs could end up being part of IBM's "value networks" just the same. If IBM is the organizer of knowledge exchange for large drugmakers, then all the partners that these big firms are allied with will appear as nodes on the client's network. Biotechs will thus also effectively become part of IBM's customer network.

As it becomes more familiar with the life sciences industry, Kovac says IBM is identifying new opportunities to be of service. The company may, for instance, pursue a "utility" model, she says, explaining, "It doesn't make sense to build an electrical plant in your back yard, but you still want to draw energy." With that image in mind, IBM may seek to sell blocks of time on high-performance computers. There's precedent in life sciences: owners of high-power x-ray machines presently rent time to start-ups that want to do x-ray crystallography but don't want to invest in that much capital infrastructure. The utility model would let IBM serve customers of different sizes: anyone who wanted access to a capacity. Even small biotech firms could send electronic files—in defined formats, through defined channels—to an IBM facility for processing. The same model is one that Cray could follow, to give researchers time on its supercomputers.

There's good rationale for outsourcing all sorts of activities in life sciences—be it high-performance computing or general IT operations, says Bob Easton. The former head of The Wilkerson Group now running his own New York-based consulting firm, Easton Associates, explains: "If you do your work in-house, it costs the same amount no matter how much you do. If you outsource, you pay according to activity. Outsourcing is good for most people for most functions, because you get to convert fixed costs to variable costs and so pay less when you do less if times are hard." Easton notes that IBM offers its services on many different pricing models, from per transaction fees to fixed fees plus gain sharing, and figures that flexibility will hold true as it moves deeper into life sciences.

What Do Drugmakers Want?

IBM is obviously hoping to leverage the reputation and experience it has earned in other sectors to give it some credibility in life sciences. So far, so good. Kovac says, "The brand is serving us well. IBM stands for reliability, security, scaleability—all the things the life sciences industry is looking for. But, she adds, "we can't come in, saying we're going to win just because we're big. We have to really understand what people in this industry need, and we think we do."

IBM certainly has a clear view of the opportunity in bioinformatics—and it is far better equipped to make the most of it than the start-ups that pioneered the sector—but the basic question is whether its potential clients see the same future. Most of the bioinformatics solutions in place at large drugmakers are home-grown—by definition customized and not standardized—in part because pharma companies perceived their quirks as a source of competitive advantage.

What drugmakers need and want, and who they think should provide it, may be quite different matters. The tension between large-scale solutions and specific applications is very much on the mind of Wes Cosand, executive director of bioinformatics at Bristol-Myers Squibb. He says, "We'd rather buy than build, but many of the applications we feel we need are not met by commercial software packages, or the packages companies are offering are too large. A lot of vendors think there's infinite price elasticity." Cosand adds that he's hesitant to put a very large system in place, when he's not sure the company will want to collect the same data or think about it the same way in six months. "We find smaller tools are more useful, and we tend to make them ourselves," he declares.

Cosand says he'd prefer having a commercial vendor assume responsibility for fixing and maintaining software, but at the same time he doesn't see a pressing need for BMS to buy a big IT package that embraces a number of different functions, technologies and databases. He explains, "We at BMS have been investing quite heavily in IT in recent years. I view us as being fairly far down the road to tying together disparate data sources and presenting to the bench scientists the bioinformatic tools they need to do their jobs." By now, many Pharma companies have invested in IT systems that reflect their corporate priorities and cultures.

It may be hard for pharmaceutical firms to let go of their old ways and learn to accept that new systems can help them more—but learn they must, argues IBM's Don Cotey. In the future, as drug research becomes far more collaborative, companies are going to need to communicate far more efficiently both inside and outside their corporate shells. They'll need data interoperability to survive and not get leapfrogged, he asserts: "If they don't adopt data standards, their legacy systems will not talk to their business partners' systems and that will kill any alliance. The CEOs get it, and the CIOs get it, but companies are still reluctant to open up the kimono."

Convincing Big Pharmas to accept and even insist on standardized protocols for storing and processing data will be one of IBM's key challenges in building its life sciences business. But it must do even more. IBM also has to convince applications developers to hew to standards as they create specific problem-solving programs. The academics and biotech firms that collaborate with large pharmaceutical firms—who have some of the same concerns about proprietary data—will also have to be persuaded to adopt the same standards. As yet, IBM has comparatively little to offer these smaller players as customers, but it is discussing making new types of data and processing capacity available on a pay-per-use basis.

Another hurdle for IBM is the fact that it, like so many discovery-based platform biotech firms, is not immediately solving the real problem in drug research: getting more compounds into and through clinical development faster. Applera's Celera Genomics Group is hardly an IBM competitor, but it began as essentially an IT effort: creating large proprietary, annotated databases and the software for manipulating them. But after bringing on a few major drug-company subscribers, Celera's big-ticket subscription sales seemed to falter—no major drug firm or biotech has signed up since Immunex Corp. , which did so in June 2000 [See Deal]. Plenty of academic groups have taken subscriptions, but the revenues from these deals are probably much lower than from corporate clients.

Market dynamics have compelled Celera to transform itself into a drug discovery and development firm: it's now applying its own IT tools and data on its own behalf, hoping to come up with products faster and more surely than the client base it once maintained it would never compete against, only support.

IBM is adamant that it is never going to compete with its clients. Doing so would be in direct opposition to CEO Lou Gerstner's mantra: "Don't compete with customers. Help them succeed by making them more efficient." That's the stated goal of the company's new life sciences business unit, but it won't be easily attained. IBM will have to get the cooperation of a decidedly ungregarious industry, from companies famously reluctant to join consortia of any kind. Even the industry-lobbying group, PhRMA, is notoriously unable to build consensus among its members on many basic issues. And when it comes to research, drug companies simply don't like to share, even on what are obviously precompetitive issues, like toxicology databases. The SNP Consortium, one exception, was inspired more by drug companies' fear that a few biotechs would corner a resource they needed than by any desire to share their discoveries.

Celera has found that it needs to show the industry its value by actually doing itself the things it originally hoped to help its clients do better. IBM is now making similar promises to help drugmakers become more efficient—but it may be uniquely suited to do so, especially if it can inspire standardization on basic data protocols. If anyone can help the life sciences industry benefit from bioinformatics and ultimately from broader, deeper application of information technology, it is likely to be the firm that best serves the sector as a one-stop shop.