This concludes a two-part report on hardware vendors' perspectives on the bioinformatics market. Part 1, which ran in the January 19 issue, discussed such customer challenges as data overload, data integration, large computing requirements, the role of specialty bioinformatics hardware providers, and the need for flexibility.
NEW YORK--While bioinformatics customers want their hardware systems to be more flexible, they're also demanding industrial-quality reliability and service, equipment vendors who spoke with BioInform agreed.
"The bioinformatics community is looking for 7 by 24 support from IT, treating it more like a production resource. Scalability and consistent service are very important, and this is not the first industry that's had this experience," said Juli Nash, biology market manager for Silicon Graphics. "These explicit trends have happened in financial markets, in automotive, in aerospace. They all got into big technical data and had to coordinate them and redesign their entire process around becoming datacentric." Now her company is applying lessons learned in those industries to bioinformatics.
"What was once, not terribly long ago, a desktop environment with a minimal reliance on serving has changed dramatically in the last two years. We're seeing the community relying very aggressively on centralized databases running on powerful servers. The distributed model of having every site do its own job is changing to one where there's a more holistic approach to designing your bioinformatics program for a global organization," she told BioInform.
"There's a very interesting trend toward commercial- and industrial-strength computer systems," agreed Lionel Binns, Digital Equipment's industry director for chemicals and pharmaceuticals in Europe. "In the past researchers wanted computer tools with power and capability, but they didn't necessarily need them on a 7 by 24 by 365 basis; it just wasn't needed. Now that requirement is moving upstream in the organization. It was formerly found in the corporate centers, then it moved into manufacturing, and then to product development areas and clinical trials. They want industrial-strength systems they can trust."
"Some sequence analysis jobs run over a week," he elaborated. "If the computer system fails, the scientists have to start over again. Today some companies are disadvantaged because they can't supply their scientists with systems of that reliability. Reliability and scalability are very, very important." He also concurred that technologies initially developed for other fields are being applied to bioinformatics. "We're looking at the application of data warehousing technologies that have been used in the commercial world to try and make sense of all the genetic data," Binns remarked.
"Visualization tools are also becoming more important," he continued, an idea echoed by several of the vendors who spoke with BioInform. Gerald McAndrew, Sun Microsystems' worldwide group manager for the chemical and pharmaceutical industries, agreed that "as time goes on you'll probably see a demand for 3-D coming back, similar to the molecular modeling days."
Nash noted that her company's MineSet product, a data mining visualization software package, "has been aggressively adopted by the pharmaceutical community. It is the largest industry that's adopted this tool, which was actually designed for financial and marketing markets for analyzing complex datasets." She added, "We support visualization technologies through sharing our visualization tools. We helped Incyte Pharmaceuticals build a visualization strategy
for their database and they now resell it."
Larry Greene, Digital's pharmaceutical industry director in the US, noted, "We've been doing some stuff in the area of visualization for years. They've used 3-D modeling on the chemical side. Now people are looking at whether they can apply it to these gene sequence data to see if they can identify potential targets earlier."
Who Makes the Decisions?
When it comes to who makes the final equipment selection decisions, the relationship between bioinformatics researchers and corporate management information systems (MIS) departments hasn't been standardized, the vendors reported. "Both sides are still learning how to skate with one another," commented Isidore Rigoutsos, who works in the bioinformatics market for IBM. "Bioinformatics people are trying to define their needs." Originally many biologists didn't have the computer knowledge to handle systems decisions, he noted, but "they're learning better and better."
"We're talking to both," Nunes confirmed. "We're talking to people who are experts in the field and who understand what we're offering and the challenges we're tackling, as well as people in the IT area who actually are the ones making some of the decisions and supplying the hardware to these people. We have a handful of customers that we perceive as leading edge where it seems to be working pretty well. Then there are others that, as Isidore said, are still trying to figure out how to communicate the needs."
McAndrew claimed, "I wouldn't say in all cases, but in most cases the big influence has been the end user, with some consulting from the IT organizations. A lot of the systems this past year have been sold directly to the end user community, the bioinformatics departments."
"The bioinformatics people by and large are extremely computer-literate," he explained. "In most of the organizations I've dealt with during the past year it's a combination. You need some really strong PhD bioinformatics-type of people heading the group, directing the computer scientists in the detail work. I'd say 98 percent of the time there's some kind of MIS representation. A lot of the IT organizations have liaison-type people."
"You tend to deal with two audiences, all on the R&D side," Greene concurred. "You tend to deal with the MIS people that support the R&D people, and you tend to deal with the scientists. But in this market, for the most part, the scientists determine the hardware platform you buy. One school of thought is, let's hire computer people and teach them the whole biological market. The other is, let's hire scientists. Most of them are very computer-literate; let them worry about the computers. And because in this market new drug discovery is everything, they can determine what platforms they want to buy. They influence it tremendously."
"Many on the IT side are very in touch with what needs to be done, although they can't go to the detail the scientists can," he continued. "An IT person at one of our customers, a biotech company, told us that after his company upgraded to our new Ultra SCSI bus, they could increase their sequencing rate to two to three times faster. And he said, you ought to share this with other customers. People are willing to help each other because this human genome project is so big. These are all very nice people and it really is a team effort."
A Big Clash
Binns agreed that the primacy of the drug discovery mission gives researchers an edge. "There's long been a split between MIS and research IT," he told BioInform. "It's a difference in requirements. As we see the requirements for reliability and scalability increasing, we may see MIS taking more of an operational hand. They're used to it. But the specification and purchase of systems for bioinformatics tends to be outside the MIS scope. Since the researchers have the ability to make the fortunes of the companies they work for, they tend to get their way." He added, "you can't generalize about the connection between corporate IT and research IT. The bigger the organization, the less interaction there will be."
But Valenta described a "big clash. It is quite common for the company IT manager to certify a certain computer brand and the scientists to want a different brand."
He said the bioinformatics market is "unlike a commercial sale, where you're dealing with the IT manager. The bioinformaticists are much more knowledgeable about computing. But the IT people are also involved in the decision-making process to some extent, because they have to integrate the systems into the company's infrastructure."
"The bioinformatics people have things they have to do, and in their priorities it is not important that the company that has the best box today may not be in business next year," Valenta continued. "The IT people absolutely do care about that, so there is a culture clash. The companies who are pure research companies, the national labs for instance, for them the main goal is to get the best box available today, because they replace their system every couple of years anyway. They don't care whom they buy from, they're going to buy whatever is the fastest at the time. A pharmaceutical or a drug design company absolutely does care."
Nash described a major bioinformatics trend of bringing science "under the roof of IT influence in a new way." She commented, "Right now the exciting thing is not only the technology of IT becoming more influential in the bioinformatics scientists' point of view, but we're seeing a closer alliance in the decision-making process, which means talking to CIO's and senior vice-presidents of IT, as well as to directors of research and of bioinformatics. There is no cookie-cutter approach; in today's world there are rare organizations where you will not at some point need to speak very closely with the scientists involved, as well as IT. The process of how those communications channels will be managed from within is still evolving."
"This is a classic environment where you have two parts of the organization working together to solve a problem," according to Nash. "Every organization is different. The consistent thing is, it's very likely you'll be working with both parts of the organization, and the maturity of that organization will affect how simple the decision-making process is."
"There's not a single organization that doesn't in some way see its bioinformatics program as extremely strategic, so we do see organizations that keep a lot of autonomy for them," she said. "But we're also seeing organizations begin to include a lot more corporate IT influence in the decisions. It's a continuum. The exciting thing is, there's a lot to be learned by sharing the technological experience with both communities. The organizations that recognize that and establish communication channels to make it happen are going to succeed."
"The technological forces of IT that were primarily held separate from research computing, the understanding of how to build integrated environments, the dependence on scalability and serviceability, and those types of key technological trends are bigger influences now in the research computing needs," Nash concluded. "How that gets articulated in an organization, whether the people actually are becoming decisionmakers, is still evolving. In some organizations the CIO's are decisionmakers. In others, absolutely not. You're seeing a great change in how the technology influence is happening."
All the vendors agreed the market for bioinformatics hardware is expanding rapidly. "We're estimating this market will grow at 75 percent a year," Greene claimed. "There are some new startups, but also there are a ton of biotech companies, and with the economy being so good, somebody said if you need venture capital, shake a tree. I would say the bulk of the market is coming from existing companies that are growing, but we had a call the other week from a small biotech startup in New Jersey. They just hired an IS guy and he wanted one of our Alpha systems."
Valenta told BioInform, "I see the market growing at a rate of more than 20 percent a year for the next five years, definitely." He noted that one of the biggest opportunities is in databases, with big rewards for "the hardware or software company that comes up with a solution to the data mining problem," adding, "pharmaceuticals are where the money is and bioinformatics is where the growth is."
"We view the whole informatics area, which includes cheminformatics and bioinformatics, as a very strong area for our company," McAndrew stated.
"There are still a lot of pharmaceutical companies that have not made significant in vestments in this area of research. There's still a sizable market left out there," he claimed, calling the education and biotech areas "relatively untapped markets."
Binns said "the potential is enormous," although he cautioned that at some point "someone somewhere in this industry has to come up with results that are realistic."
In response to perceived opportunities in the market, many hardware companies have mounted special bioinformatics efforts. For instance, "Silicon Graphics has been supporting the bioinformatics community for two and a half years," Nash said. "We now have four or five dedicated, 100 percent staff supporting bioinformatics. That includes industry management like myself, applications engineers, and technical support. We do projects with the community. We've done a number of what I would call big compute projects, as well as projects that have pushed the envelope on visualization and delivery of information to the community. One of our first projects was two years ago, with the European Bioinformatics Institute, where we supported the first annotation of a eukaryotic genome."
At IBM, Nunes said, "we see this as a growing opportunity that plays into multiple IBM strengths, and the fact that we have leading-edge research in the biotech area--and bioinformatics is one piece of that--helps us to understand what our customers are facing. We have some customer partnerships that are going on to help us even more intimately understand what their challenges are."
"We've had a few people working kind of underground in this area for four years," she reported. "We just decided in the last 12 months that this should be one of the key projects for IBM research in 1998, and in that light we put together within research what we call a computational biology center. We've actually made that a group that is pulling on the resources of the people in Zurich, Switzerland, and Almaden, Calif., and folks at the Tokyo research lab, so that we now have what we hope in 1998 will become an integrated effort worldwide. It's about 10 people, which, for a research project, is large; they are often three or four people. And we hope to grow that modestly in 1998. We're working very closely with the pharmaceutical people because a lot of the customers we would be talking with about these solutions are in that industry."
At Digital, Greene observed, "We as a company are trying to think through what we call a bioinformatics infrastructure. What are the components of it? And then, what can you do to target the data in a way that you can identify early targets faster?"
What the Future Holds
All the companies contacted by BioInform promised significant new products with relevance to the bio informatics market in upcoming months, although some were more willing than others to give details. Greene ticked off Digital's list: "We just introduced the new Ultra SCSI subsystem array. The 625 MHz chip will be coming out in volume. And then we have a multiprocessor system, I think it will be just a little bit past six months, that we're doing nondisclosures on that they're very excited about."
At Sun, McAndrew said, "there are a lot of areas we can take a look at to improve the machine architectures that would enhance the whole informatics area." As an example he cited a French supercomputing lab that discovered "if they rewrote the Smith-Waterman algorithm using the hardware integer instructions within VIS, the visual instruction set in our UltraSPARC architecture, they got almost a two-time improvement in performance without doing anything else."
According to Nunes, IBM "has tools for data mining, we have some data storage and data management solutions that were already in IBM's product portfolio; there are lots of pieces that we have existing as well as some things in the research pipeline that we hope to be bringing out in the next 12 to 24 months. Nothing is definite as far as product lines, but we're working with our partners in industry to look at the right time to bring them to market, sooner rather than later." IBM's first "research project aimed specifically" at bioinformatics is focused on a data mining product that has been optimized to be much faster and more efficient in bioinformatics applications, Rigoutsos revealed. Nunes added, "We think we have an advantage over products that were not developed specifically for this arena." IBM hopes to bring that product to market this year, she said.
Finally, Nash reported, "People come to Silicon Graphics because they need big servers to support complex data, big databases, they need 64-bit performance, lots of robust architecture, but the desktop environment is, in this community, Macs and PC's. We will be supporting a product in that range in the second half of 1998. We don't have any announcements about it except that it will be a Wintel product."
"The reason I'm particularly excited about it," she continued, "is that this is exactly what the bioinformatics community is looking for: big Unix server commitment for the technical applications, as well as being able to support the end user environment, from a single distributor. We found that many decisions were being made increasingly higher in the organization. They involve CIO's, whose total environment includes not only the Unix serving environment but also supporting all the desktops. Now we will no longer have to bring in third parties to satisfy that entire range of needs."
"We're no longer going to be just in the server environment. We're going to get to provide all that complex computing environment, as well as the desktop. In the future our plan is, when we talk to the CIO we'll get that entire purchase order," Nash said.