Ann M. Thayer

C&EN Houston

Drug research is data rich but information poor. Genomics or gene-sequencing projects, high-throughput screening, combinatorial chemical synthesis, gene-expression investigations, and pharmacogenomics and proteomics studies are creating massive volumes and multiple sources of biological and chemical data. Data threaten to cause a bottleneck in drug discovery and development. Relating and turning this complexity of data into useful information and knowledge is the primary goal of bioinformatics.

As computing and biology have converged, software tools for data capture, management, analysis, mining, and dissemination have emerged. More than 40 companies, most of them small, are trying to capitalize on the development and marketing of new bioinformatics tools. Whereas the market for generated data or "content" is very lucrative, bioinformatics sales are expected to reach about $160 million this year, according to market research firm Frost & Sullivan , Mountain View, Calif.

Consultants and companies envision tremendous market potential—as much as $2.0 billion to $2.5 billion in five years—if drug and biotechnology companies together spend on the outside as much as they've been estimated to spend internally on bioinformatics. The need for new, faster, and more sophisticated tools is unequivocal. And consultants predict double-digit rates of sales growth in a market that is less than saturated.

To successfully map the genomes of various species, Celera has begun operating what will be one of the world's largest data centers. [Courtesy of PE Corp.]

But low market saturation, say many in the business, may be more a problem of accessibility than finding the untapped buyers. Commercial bioinformatics software typically is very high priced, with larger packages selling for tens or hundreds of thousands of dollars. Customized systems can cost millions of dollars, not including computer hardware. On yet another level, academics and other researchers are giving away software.

It's not surprising, then, that most customers have been large pharmaceutical companies that can afford bioinformatics tools and don't want to depend on giveaways. However, these same customers also can be competitors in that many have extensive in-house bioinformatics efforts under way. Although their need is great and they are willing to spend R&D dollars on bioinformatics, major pharmaceutical firms still represent a limited market.

In addition, software providers have designed many bioinformatics tools with expert users—bioinformaticists or bioinformaticians, as they are called—in mind. But there is a broader market, bioinformatics companies believe, comprising researchers who want and need a variety of easy-to-use tools. Many work within the large corporations, but even more are in smaller companies and universities with limited resources. Removing the barriers to top-notch tools and reaching a larger customer base is a goal of several bioinformatics companies.

"Who I really want as customers are the scientists themselves," says John Couch, chief executive officer of DoubleTwist , Oakland, Calif. "We're challenged in this field to deliver something to the scientists so that they can do their science." In December 1999, Pangea Systems, one of the more successful of the bioinformatics tools providers, recast itself as DoubleTwist.

Bioinformatics dot com

"What this industry needs is an environment that reduces the cost and complexity of genetic research," Couch continues. As an applications service provider (ASP), DoubleTwist operates an Internet-based "research portal." Through the portal, researchers can access tools (many originally developed by Pangea) and search algorithms, databases, patent and scientific literature, news, and jobs listings. Automated agents take requests and then retrieve and interpret data, returning results to the user.

Bioinformatics can combine data from many sources Gene discovery High-throughput genetic sequencing Genetic algorithm searches Genetic linkage studies Genetic maps Polymorphisms (SNPs) Gene function Gene chips and microarrays Gene expression profiles Functional genomics Proteomics Metabolic pathways Drug design Computer-aided drug design and molecular modeling High-throughput screening Structure-based drug design Small-molecule-compound libraries Virtual chemical libraries Drug development Clinical trial data, efficacy data Pharmacokinetics Pharmacogenomics Pharmacogenetics Toxicology Patient data and medical informatics

The site will offer three subscription levels. Bronze, which is free, will provide limited access to some databases and search tools and should serve the academic community, Couch says. Silver and gold levels get additional sequences and proprietary databases and tools. Low-level access may cost about as much as a journal subscription, or about $500 per year, Couch estimates, whereas commercial subscriptions could run a few thousand dollars per "seat," and "full-fledged genomic analyses" may be more expensive.

DoubleTwist has been testing its site at Stanford University and with the biopharmaceutical firm Chiron, Emeryville, Calif., and small drug discovery company Tularik, South San Francisco, Calif. In mid-January, a trial version of the site was opened to the public. DoubleTwist recently announced that it has signed on 13 new partners, for a total of 20, for data content, tools, computing infrastructure, and e-commerce.

Combining the Internet and bioinformatics is not new. Major public databases—such as GenBank, maintained by the National Center for Biotechnology Information , and others from the European Molecular Biology Laboratory (EMBL) and European Bioinformatics Institute (EBI) —have long been available through this channel. And DoubleTwist has not entered the Internet-based ASP market alone.

eBioinformatics' BioNavigator Internet site is a spin-off from efforts dating back about nine years at the University of Sydney and the Australian National Genomic Information Service . The Pleasanton, Calif.-based ASP was created about two years ago to make low-cost and easy-to-use bioinformatics tools and resources available to researchers in academia and in small and midsized biotechnology and drug discovery companies.

As of Jan. 22, eBioinformatics dropped its subscription model, opting instead for a pay-as-you-go system. Registered users can run certain gene-analysis programs and store data for free. "Only when they want to go to the next step in the process do they then begin to burn 'eBio' units," explains James Nelson, vice president for product marketing. The billing schedule is based on the kind of program run and how much computing power is required. "A relatively active molecular biologist might run about 1,000 units per year," he notes, "and at that level the units are about 50 cents each."

Nelson emphasizes that BioNavigator has licensed many popular algorithms and programs for gene and protein analysis through more than 20 alliances. As part of its services, it provides private work spaces where user data can be stored and analyzed on its server rather than being transported back and forth across the Internet. To work in a group or team, users can identify other registered users who can share data.

Security and privacy are important concerns. Nelson points out that data are transmitted securely only once and maintained in a secure environment. eBioinformatics agrees never to access data except under user direction, and it will not share information with other customers. The company also does not track what kind of data is stored or what analysis tools are being used, which can be a competitive concern for drug discovery firms.

Viaken Systems believes it can offer similar capabilities to small and midsized companies in a more secure fashion than through a public Internet site. The Gaithersburg, Md.-based ASP provides and manages complete bioinformatics platforms, including software licensing, hardware, training, and support, on an externally hosted server. Clients link through virtual private networks to user-dedicated hardware, data storage, processing, and e-commerce capabilities.

"Because of the tremendously valuable information that can be the 'crown jewels' of most of these companies, we believe that security is a critical element of providing a full-service environment," says James W. Serum, Viaken's chief operating officer. "Because of the high-speed connections and secure data links, the outsourced bioinformatics service behaves essentially as if it's inside their organizations."

Last year, Oxford Molecular's Genetics Computer Group (GCG) granted Viaken an exclusive license to offer GCG's bioinformatics products through a hosted environment. GCG's products include the Wisconsin package, one of the most widely used and comprehensive sequence-analysis tools available.

ASPs believe they have a competitive advantage in being able to continually update data and tools. "Software needs change, and a single software package rarely meets a scientist's needs over any length of time," Serum says. "The rate of change of technology and capability is so fast that users can't keep up with it alone or afford the large capital investment. So we make alliances with a wide variety of best-in-class and public software capabilities to give users a complete portfolio."

Viaken also offers high-performance computing, which requires a significant capital investment to install and support. Viaken spreads the cost over multiple users and charges on a per-usage basis, which makes it affordable for small or infrequent users. Viaken and Hewlett-Packard together are developing hosted high-performance computing solutions. HP also chose Viaken as its preferred ASP for the life sciences market.

"In the past, software and hardware vendors have focused only on the very top tier of pharmaceutical companies," Serum says. But small, midsized, and even large companies face similar challenges and have similar computing needs. "Beyond the critical few big pharmaceutical firms, you have a world of smaller companies that are generating huge amounts of data and need access to these tools."

Compugen , Jamesburg, N.J., has been known for its software and hardware, having supplied bioinformatics systems for gene searching to a few major drug companies and the U.S. Patent & Trademark Office. In December 1999, it launched its LabOnWeb Internet-based research engine that is still in the testing phase. Test sites include Harvard; New York University; the University of California, San Francisco; Hebrew University in Jerusalem; Tel Aviv University; and the Weizmann Institute of Science, Rehovot, Israel.

Based on user feedback, the site continues to evolve in content and appearance, says Lior D. Ma'ayan, Compugen's chief operating officer. Although not finalized, LabOnWeb probably will include subscription-based plans for commercial organizations as well as pricing per query or information request. "If you package it right, the web enables you to reach many more people and actually get to the bench scientists," Ma'ayan comments. "And if you do it right, you actually deliver real value in a way that is the most efficient."

One-stop shopping

To facilitate the research process, many Internet-based sites and ASPs are including e-commerce features such as links to suppliers of reagents, DNA sequences, or research clones. DoubleTwist's Couch calls it "contextual e-commerce" in which suggestions or links to suppliers can be returned with data search results. DoubleTwist recently hooked up with SciQuest.com to let users order research supplies from its electronic marketplace.

Content providers are using the Internet as a means to sell gene sequences and their associated bioinformatics tools as well. The Institute for Genomic Research (TIGR) , Rockville, Md., makes its databases and a number of tools available free for researchers at nonprofit institutions using them for noncommercial purposes. Geneva Bioinformatics , exclusive licensee of the Swiss Institute of Bioinformatics (SIB) , offers the same deal for access to SIB's Swiss-Prot protein database. However, commercial users can spend a few thousand dollars per user or up to $90,000 per year for unlimited access.

In May 1999, Curagen , a New Haven, Conn.-based genomics and drug discovery company, launched an Internet portal called GeneScape . It gives access to the company's CuraTools gene- and protein-analysis software and selected proprietary databases, and it will give access to public sources such as the Human Genome Project and the SNP Consortium database of single-nucleotide polymorphisms or genetic variations. A freely accessible version is intended primarily for academic use, and commercial users get a 30-day trial period to evaluate the site.

In contrast, Incyte Pharmaceuticals , Palo Alto, Calif., has been more exclusive, garnering multi-million-dollar licensing fees from most of the major drug and biotech companies that want to see its gene-sequence databases. Its partners can also use a collection of bioinformatics tools, called LifeTools, that are available through the company's web site. The software processes, stores, analyzes, and manages both proprietary and public genomic data and includes some project management and data sharing for research teams.

Gene Logic , Gaithersburg, Md., combines data management and analysis tools with its gene-expression databases. Because Gene Logic's data are warehoused in electronic form and in a relational database, the company has the flexibility to repackage and price the data in different versions. In the past few years, it has been constructing custom databases for several major drug partners.

The company just recently signed on its first multiyear subscribers to its new large-scale GeneExpress database. Three-year subscriptions are between $3 million and $5 million per year. GeneExpress users connect over secure Internet links to Gene Logic's main computer facility. Through this web interface, they can search and analyze the database, saving results in dedicated work spaces.

The various public and private databases often have different and incompatible structures. Gene Logic emphasizes that its "object-oriented" software lets customers manage, integrate, and query these databases as if they were part of a single database. And, it adds, because the information is distributed over the Internet, the company envisions future money-making e-commerce opportunities with a portal system.

Hyseq , Sunnyvale, Calif., has taken a different approach, one that it hopes will help it capture a broader segment of the market. Rather than requiring multi-million-dollar fees for access to its databases, Hyseq has made available to all researchers its proprietary gene data, which can be purchased on a per-item basis. Last fall, it launched GeneSolutions . com as the vehicle for gene-sequence, homology, and expression data.

Similarly, Lexicon Genetics' Lexgen . com site offers free searches of a portion of the company's OmniBank gene database that represents known human and mouse genes. Corporate subscribers and academic collaborators have access to new discoveries in OmniBank. Once users find a desired gene, the Woodlands, Texas-based company hopes to sell them "knockout" mice engineered for gene-expression studies.

Although it produces content, Celera Genomics, Rockville, Md., the gene-sequencing powerhouse that is part of PE Corp. , calls itself an information, rather than a genomics, company. Its Internet-based business model includes a strategy for delivering genomic information electronically to the biomedical and agricultural research communities. Revenues come from database subscribers, but to promote use of its data, the subscribers, not Celera, have rights to any discoveries made using the data.

Celera already has drug industry subscribers that pay $5 million per year for five years. Celera has said it expects to extend subscriptions to academic and other researchers at a different cost this year ( C&EN, Jan. 17, page 11 ). Although its Internet site for genomic data is not yet operating, Celera says it intends to offer data-mining, analysis, and visualization tools as well.

Uniting the R&D masses

In the same way that individual users and small companies may move to integrated solutions on the Internet, large pharmaceutical firms are seeking ways to integrate their own drug discovery and development processes. They want to tie together separate R&D operations—gene discovery and function; target screening and selection; chemical structure and synthesis; and development efforts in toxicology, pharmacogenomics, and clinical studies—and deliver the information and tools across an organization to individual scientists on their desktops. Companies want to do this without losing their investment in proprietary or legacy systems.

"Although we certainly still need them, having newer algorithms isn't really the biggest problem, [it's not] where the bottleneck is right now, [and it's not] where there is value to a pharmaceutical company going forward," says Steve Gardner, CEO of Synomics , Cambridge, England. "In the postgenomic era, we're going to have all this information. Getting the data into the right form and to the right people and allowing them to look at it in the right way with all the other information that's relevant in decision-making is really the challenge."

Data volume alone also isn't the issue because "you can just throw computing power at the problem," Gardner believes. "What really does get difficult is that you're expecting a user to sit in the middle of all this growing information and deal with the much more complex relationships that we have now between data. It's a terribly intimidating place to be."

Differing data types and formats, software languages, operating systems, and computer hardware complicate integration efforts. Integration typically is accomplished by creating small, object-oriented software elements, or "wrappers," that let a single overlaying, often browser-like, desktop application interact with all the pieces. The original separate systems are intact and functional, and new ones can be added, while the underlying complexity is transparent to users.

There still are many challenges and different degrees of success in integration at the level of both data and processes, but computing environments are becoming more unified, flexible, and expandable. To make the process easier, informatics and drug firms have set up a life sciences research task force within the international Object Management Group to establish software and computing standards. There also are several projects in ontology to create definitions or descriptions of biological data types.

Companies want to integrate processes and information because they can generate a competitive advantage by making better decisions, based on all available information, and make the decisions faster, Gardner explains. Synomics is working with a number of top-20 drug companies on projects that range from smaller to million-dollar-plus deals, he says. Projects often include R&D processes and data—even bibliographic sources, competitor information, and patent data—as well as everyday work tools such as word processing and spreadsheet programs.

Synomics offers integration architecture, called Alliance, that lets public, proprietary, and third-party licensed databases, processes, software, and applications be accessed simultaneously. "It manages the complexity and the difficulties of dealing with that," Gardner says. "Then you can build on it with data visualization and project management tools to allow all of those technologies that you invested in to work together better and be put in the hands of people in an effective way." To this end, Synomics created a desktop interface called Project Explorer.

Base4 , Carlsbad, Calif., recently announced that, after one year of marketing, it has orders from more than 12 major pharmaceutical and biotechnology firms for its PharMatrix knowledge management system. PharMatrix provides project support with access to internal and external databases and analysis tools. The system can be used, for example, to capture and disseminate data and information involved in target identification, assay development, screening and lead optimization, and preclinical development.

"Future success depends on how good a job a company does now in selecting targets, developing assays, optimizing new drug candidates, and getting them to the clinic before its competitors," explains Martin Sumner-Smith, Base4's president and CEO. "There is a huge amount of unstructured, ad hoc data around the decision to pick one target versus anything else, and that requires a knowledge management system," he adds.

"The two biggest issues for pharmaceutical companies are knowledge management and the discovery/development interface," he continues. "In many ways, those are the same thing. The problem with the discovery/development interface is that what people know in the discovery phase isn't being passed on to the people doing development." Knowledge management, he suggests, "is changing this linear development process into an iterative one."

Other companies, such as LION Bioscience , Genomica , and NetGenics , have developed similar concepts. Heidelberg-based LION (Laboratories for the Investigation of Nucleotide Sequences) has several tools that are components of its "i-biology" integrated environment. For example, its SRS6 software, licensed from EMBL and to several major drug firms, integrates disparate databases through a single query and navigation interface, allowing for simultaneous or cross-database searches. LION also produces bioScout, an automated gene-analysis tool.

LION, which also conducts genomics work, is probably best known for having signed a $100 million deal with Bayer in June 1999. Under their five-year agreement, a LION subsidiary is creating an information technology management system for drug discovery.

Genomica, based in Boulder, Colo., has a system called Discovery Manager, which manages software tools, databases, and projects. According to the company, the object-oriented Discovery Manager software will incorporate results from epidemiological studies and clinical trials into the same database structure as the genomics data to allow for data sharing and querying. The software has been licensed to both Glaxo Wellcome and the National Cancer Institute.

Before it became DoubleTwist, Pangea had created the Pangea Unified Life Science Environment, or PULSE, for integrating and analyzing genomics and biological information. Although PULSE is compatible with different computer platforms, it changes the underlying structure of public and proprietary databases and consolidates them into a centralized repository. It also includes tools for gene-sequencing analysis and project management.

NetGenics' Synergy is another enterprise-wide framework that supports different databases and applications. Last year, the Cleveland-based company released Distributed Synergy, which enables real-time sharing of data, project information, and analytical functions between research teams at multiple company locations. It has deals with Pfizer and American Home Products.

"People tend to forget that, in addition to integrating data sets and tools, the third thing that needs to be integrated is the actual people doing the work," says Manuel J. Glynias, president and CEO of NetGenics. Synergy therefore incorporates what Glynias calls "teamware."

"We want the organic chemist, the molecular biologist, the biochemist, the enzymologist, the medicinal chemist, and other scientists all to be able to work together to solve problems," he explains. "They all have access to each other's data, not at the depth that only a specialist would understand, but at a summary level where it is interesting and useful to everybody. That's a real missing link in this sort of software and something we've been trying to build."

Data, data everywhere

NetGenics' approach is to leave data in independent databases that can each be accessed through one channel. Users can be unconcerned with a tool's or data's source, format, or location. However, existing instrumentation and software applications can operate unchanged, databases can be updated, and specialists can still work with familiar or specific data for more detailed analysis.

NetGenics recently agreed to help IBM market its new DiscoveryLink product, the first from IBM's life sciences organization formed last year. DiscoveryLink creates a virtual database as though the data were in a single database. The data remain highly heterogeneous in format and decentralized. The alternative, Glynias explains, is converting all the data into a common format and storing it in a data warehouse.

Silicon Graphics' Visualizer MineSet scatter plots let researchers visually examine data relative to parameters. The vertical purity axis indicates signal consistency for each gene over the 20 features measured. The avg diff axis indicates comparative expression levels (genes toward the front are expressed at higher levels). The horizontal axis, chgf, indicates the change observed in expression level (points further to the right or left indicate genes that have been down- or up-regulated, respectively). Blue and red points indicate genes that have been turned off or on, respectively. The size of the point indicates the reproducibility of the experiment. [Courtesy of Roche ]

"DiscoveryLink addresses a core issue: the ability to perform complex queries—queries not even possible today—across heterogeneous data sources," Glynias says. DiscoveryLink understands the schema of different databases and what kind of queries each can handle, he explains. Then, when a person or program poses a query, DiscoveryLink breaks it down and sends the parts off to the various databases. The partial answers are combined, and an answer is returned.

PE Informatics , a division of PE Corp.'s PE Biosystems Group , has a software product called BioMerge. At its core is a relational database that lets users integrate and query disparate proprietary, public, and other third-party data for analysis and annotation. It uses technology from Oracle, which produces some of the most widely and generally used information management tools.

Data can also be exchanged securely among BioMerge servers; for example, Celera's genomic data can be delivered by this route. And PE Informatics sells a software suite called Discovery Tools that works on top of BioMerge for data analysis, and other programs, such as GeneKeeper, for data management. The company anticipates expanding the capabilities of BioMerge as revisions are released.

"The focus of the market right now is still largely on the [gene] discovery end," says Catherine Baldwin, marketing manager for PE Informatics. "We believe the next big step is proteomics, and gene-expression data are part and parcel of that. After that, people are going to want [to work with] chemical structure and metabolic pathway data," she notes.

In data management, the first steps have been accessing separate databases from one application followed by more sophisticated data warehouses or virtual databases for advanced querying. However, sometimes the challenge "just is getting some types of data into an electronic format," comments Edward Hodgkin, senior director for contract and discovery research at Tripos. "A great deal of information has not been available to discovery scientists."

Through its software consulting services, St. Louis-based Tripos provides integration services for biological, chemical, and other data to major companies. One of Tripos' major integration products is its MetaLayer architecture. Tripos and other companies that have strengths in chemical informatics—such as Oxford Molecular with its Diva program, MDL Information Systems' Chemscape, and Molecular Simulations' WebLab—are already producing tools that combine chemical library, lead identification, physical property, and molecular structure data.

"When scientists and companies see the power of data integration and disseminating information across an organization, the imperative to get data into an electronic form will be understood," Hodgkin continues. "The integration of data allows one to make decisions and that must feed back into the design of experiments. So, the next step is to use data integration to drive experiments in a seamless fashion."

Seeing is believing

Having data in an electronic format is worthwhile only if scientists can actually work with it in a way that is useful. "It's critical that scientists have the ability to visualize data and look for trends because visualization is the mechanism for extracting information," Hodgkin says. Data visualization consists of graphs, images, spreadsheets, molecular modeling, and any other ways of presenting and interpreting information.

Spotfire , Cambridge, Mass., has taken the data visualization market by storm. The company says that it has the top 25 pharmaceutical and 40 major biotech and ag biotechnology firms in the U.S. and Europe among its 200 customers. Bioinformatics partners also incorporate Spotfire's products as the visualization tools in their overall data management and analysis.

Spotfire's lead product is Spotfire Pro, with which users can interactively mine or explore multiple or massive databases and construct visual representations of the data. Its Decision Explorer add-on uses algorithms to assist users with integrating, navigating, and analyzing complex data sets to show correlations among desired variables. Another extension to the program is its Discovery Server, which lets researchers put the visualization results in electronic reports that can be shared with others.

Spotfire also produces other modular software packages such as Structure Visualizer, which can link Spotfire Pro with MDL's ISIS program, a chemical structure database tool, or Oxford Molecular's RS3 Discovery database of biological activity and chemical structure data.

Visualization tools for the drug research community also are being developed by major computing companies such as IBM and Silicon Graphics ; smaller companies such as Neomorphic, Berkeley, Calif., and Silicon Genetics , San Carlos, Calif.; and various academic and industrial consortia.

Silicon Graphics, Mountain View, Calif., has data-mining and visualization software called MineSet. The company also is a major supplier of computers, software, and algorithms to the bioinformatics and cheminformatics markets. MineSet is being used at EBI and at drug companies including Zeneca Pharmaceuticals (now part of AstraZeneca ), Roche , and Bayer with gene-sequence, proteomic, protein and chemical structure, high-throughput screening, and gene-expression data.

Likewise, Neomorphic has collaborated with University of California at Berkeley, Novartis , TIGR, Genentech , Monsanto , and SmithKline Beecham . The company has helped build what will be the front-end browser application for Celera's genomic databases. The browser ties together large amounts of disparate information, Neomorphic says, into "an easy-to-navigate pictorial gene landscape, allowing scientists to study specific genes in the context of a complete genome."

In July 1999, Silicon Genetics launched GenEx , a public Internet database for gene-expression data from microarrays, DNA chips, and related technologies. Using features from Silicon Genetics' more comprehensive genomics and proteomics visualization and analysis program called GeneSpring, GenEx is designed to let researchers publish text and image files on the Internet in a format that is viewable by other scientists using any web browser. Silicon Genetics is providing free access and plans to work with the National Center for Genome Resources to expand GenEx' capabilities.

Whether it is with data visualization programs or through integration efforts at large corporations and Internet-based ASPs, the goal increasingly is to put the tools in researchers' hands. Bioinformatics offerings continue to evolve and target individual scientists, often through their desktop computers.

Competition is heating up, say company executives. "But it's actually very good news because it's a sign that it's about time to deliver bioinformatics to the bench scientist," Compugen's Ma'ayan says, and others agree. "Historically, informatics knowledge was in the hands of a few, and now it's time to deliver it to the masses."

Bioinformatics companies say they will increase access to more data and tools as they are generated, moving beyond the most widely used gene-sequencing analysis to include areas such as gene expression, protein identification and structure, biochemical pathway data, pharmacogenomics, and chemical structure and activity. From a user perspective, scientists hope to get integrated packages of data, software, patent citations, literature, and supplier links to support their research. These combined elements are anticipated to decrease time spent handling, manipulating, transmitting, and analyzing data to ultimately speed up drug discovery and development.

Chemical & Engineering News
Copyright © 2000 American Chemical Society