GAMING

By Steve Conway, Bioinformatics Business Director, Cray Inc. -- Remember the Internet market? I mean the land rush market of yesteryear, not today’s more modest, civilized sector. Some of the same risky dynamics that caused that early, Wild West market to falter now threaten to slow scientific progress and financial growth in the nascent bioinformatics (a.k.a. life sciences) market. Associated Press reporter Leslie Gornstein chronicled the sector’s growing pains in July 2001: “A year after the deciphering of the human genome boggled the world, investors are realizing that [bioinformatics] is still in its infancy - and far from profitable. The goal is to develop new drugs more quickly. Last fall, [bioinformatics] companies were shining stars in an otherwise dead market for public offerings. But financial analysts say the companies are no closer now to unlocking genetic secrets than they were then. As a result, stock prices have plummeted, venture capital is slowing and an industry consolidation has begun.” Things have picked up since July, but the underlying issues remain. High on the list of risks shared by the early Internet and bioinformatics markets are hyperbolic marketing and inattention to “fundamentals.” For too many Internet firms, the fundamentals of operating a business were lacking. In the bioinformatics sector, the frequent culprit is inadequate scientific rigor—a constraint imposed by the imperfect state of biological knowledge, and compounded by the limitations of today’s mainstream IT solutions. As one noted researcher described it to me, “All too often, current methods require guesswork on top of guesswork.” Keep in mind that today’s typical problems are elementary when compared with what will be needed for the confident creation of new drugs and other medical treatments: a profound understanding of the sequential process of protein-folding, of metabolic pathways, cell and organ function, and so on. This exciting field, barely out of the starting gate, is already hitting tall IT speed bumps. To complete problems in reasonable timeframes, scientific insight and problem resolution are often sacrificed. The heuristic pass at the front end of standard BLAST is a good example. This shortcut typically is used because running the more rigorous Smith-Waterman algorithm would take too long. The “poor prospects” that are filtered out by the BLAST front end aren’t necessarily unimportant. In fact, these low-scoring discards may contain indispensable information for scientific insight and drug design. It is possible that organizations relying too heavily on standard BLAST are doomed to lag behind more-farsighted scientific and business competitors. Then there’s all that “space junk,” the 85 to 90 percent of the genome that’s non-coding and therefore assumed to be inactive. Finally, what about the continuing need to understand the relationships between genomic and proteomic information? In the current debate, proteomics enthusiasts too often dismiss genomics work as passé, an early preoccupation that is irrelevant for drug discovery and design. But how is it possible to understand the origins of diseases, for therapeutic purposes, without understanding why these diseases are associated with specific locations (“microsatellites”) on the Human Genome? For the health of this market, genomics and proteomics work ought to advance hand in hand. Cray’s mission in the life sciences is to make rigorous, first principles methods fast and affordable enough to apply in all cases. We provide “brute force” computing power to leading organizations, and collaborate with them in applying it. The benefits are faster, higher-fidelity scientific results; more efficacious drugs and treatments; and stronger competitive advantage. Beyond that, we hope that our customers’ successes will contribute to a healthier life sciences market. Our predecessor firm, Cray Research, was a co-sponsor with the U.S. Department of Energy of what arguably was the first global bioinformatics conference, held in 1990 at Florida State University. The market didn’t gain momentum until mid-2000, when the Human Genome was largely sequenced. We began getting serious in May 2001, after the National Cancer Institute validated that special hardware built into our supercomputers for the intelligence community handles sequencing tasks more than 100 times faster on average than other available solutions, and does this exhaustively, with absolute rigor. Well before this, BioNumerik Pharmaceuticals in San Antonio began using “brute force” capabilities on Cray supercomputers to support drug design from first principles, including large-scale protein-folding far in advance of energy minimization approaches. For the life sciences market, we have a multidisciplinary global team. In early 2002, we plan to form a bioinformatics advisory group of leading customer and research organizations to help direct our future path in this market. Because Cray is new to this market, it’s been very encouraging to see how quickly interest has grown in collaborating with us. We operate in close partnership with each of our customers, and are currently discussing collaborations with leading life sciences organizations in the U.S., Europe, Asia and Africa. We welcome inquiries about research collaborations on challenging problems of scientific and economic importance. Contact Jef Dawson, Bioinformatics Marketing & Development Director (jef@cray.com). The Cray SV1ex™ supercomputer system, with special hardware and software capabilities for bioinformatics, began shipping in November 2001 as the newest addition to our product line. We also offer microprocessor-based SMP clustered solutions, along with technical services. For organizations needing extreme performance, the Cray SV2™ will join the Cray SV1ex in the second half of 2002. The outlook in the HPC segment of bioinformatics looks bright at the moment. Bio sales of HPC systems are becoming brisk. Dollar estimates for sales growth vary greatly, but all are in the Carl Sagan range (“billions and billions”). Clearly, the long-term prospects for the supercomputing segment are wedded to the health of the overall bioinformatics market. If bioinformatics methods do not produce an adequate flow of new drugs more quickly and assuredly than traditional methods, the future for HPC bio-solutions will be dimmer than projected. Cray looks forward to collaborating with customer and research organizations to enable the kind of scientific advances needed to realize the market’s exciting potential.