Chasing the genetic code in an IBM world

By Amy Harmon

New York Times News Service

HAWTHORNE, N.Y. -- Gustavo Stolovitzky, a brash biophysicist at the IBM division conducting basic genetic research, seems an unlikely catalyst for a multimillion-dollar business investment.

"Revenue, marketing, mind share, I don't care about these things," Stolovitzky said recently. "I care about my scientific interactions, my knowledge, my number of papers."

Stolovitzky might not look at income statements, but IBM is counting on researchers like him to give the company an edge in selling products and services tailored for science and medicine.

Stolovitzky's research collaboration with a Mayo Clinic biologist has already helped spawn a project for IBM to digitize the genetic profiles in millions of the clinic's records.

That is the kind of happy accident IBM is aiming to replicate. Whether the 150 Ph.Ds IBM employs to study life sciences in its research labs care about business (mostly they do not), they are playing an increasingly important role in it as the company woos a new kind of customer born from the genomics revolution.

IBM, which is based in Armonk, N.Y., is pursuing those customers -- scientists, biotechnology and pharmaceutical companies, hospitals, universities and government health departments -- with projects like the oft-touted Blue Gene, a $100 million supercomputer that is being designed for complex biological simulations, to InsightLink, which is new software that translates paper-based research methods into a searchable online system.

"This is a change for IBM," said Carol Kovac, a chemist by training who started IBM's 3-year-old life-sciences business unit after serving as a vice president for research.

"We're talking to scientists at the bench, the head of informatics, the head of R&D.; Linking into that community can only be done by people who truly understand it."

Opportunity for growth

At a time when information technology sales are sluggish in areas such as financial services and retail, the life-sciences business will grow at bearly 20 percent over the next three years, according to analysts. Along with its competitors, IBM is investing in it heavily.

Now in the lead with 15 percent of the $14.6 billion life-sciences market, the company has a slight edge over its nearest competitor, the Hewlett-Packard Co., according to IDC, a computing industry market-research firm. And IBM hopes to exploit the expertise and professional connections in its research division to give it an added lift.

About 5 percent of IBM's research division is devoted to life sciences, although the fledgling life-sciences business unit brings in far less than 5 percent of the company's revenue.

Analysts say revenues for the new business division are expected to reach close to $1 billion this year.

"Developing software for scientists is a completely different game than developing software like Microsoft Office," said Nathan Cosper, an analyst at the consulting firm Frost &; Sullivan. "IBM's research efforts give them more credibility in the area than any of their competition."

Credibility, of course, does not necessarily translate into business. While biologists are increasingly relying on computers to sift through reams of genetic information and model processes in silicon that their instruments do not allow them to observe in reality, IBM's offerings are often more expensive than other options.

At a conference earlier this month co-sponsored by IBM and Brookhaven National Laboratory to ask scientists how they would like to use a Blue Gene supercomputer, some said they would rather spend the money on their own technology.

Many who run intensive simulations, for example, pass up packages from IBM and other name-brand vendors for cheaper groups of linked computers known as Beowulf clusters. Such systems are often cheaper than comparable mainframe computers.

"You need to ask yourself: What's the most cost-effective way to do the research?" said Terrence Sejnowski, a computational biologist at the Salk Institute for Biological Studies, who uses a Beowulf cluster to model how synapses work at the molecular level. He acknowledged that, when it is completed, Blue Gene will run about 650 times faster than his cluster, but since he would likely be able to pay for access to only a small fraction of Blue Gene's power, Sejnowski said he would prefer to use lesser computers he could run 24 hours a day. "Except for a few cases," he said, "the trade-off almost always is it's cheaper to run commodity computers and wait until tomorrow to get your results."

Finding questions

Still, Sejnowski and many other biologists say that science would undoubtedly benefit from Blue Gene, whose speed will dwarf any of today's supercomputers.

In addition to modeling protein folding, thought to be essential to understanding how to treat many diseases, the software written for the supercomputer will seep down into future generations of slower computers, benefiting a wider range of life-science projects.

"Some of it is going to be serendipity," said Carl Anderson, chairman of the biology department at Brookhaven. "By having a machine like this you will find that there were questions no one had thought of asking."

Paul Coteus, the program manager for Blue Gene, said the biologists in IBM's research group were already lining up to use the sliver of the machine that his group had assembled in the last two months after receiving the first of what would eventually be 65,500 specially designed chips. An important test of 512 of the assembled chips at the company's Thomas J. Watson Laboratory in Yorktown, N.Y., this fall will determine whether the design works as envisioned.

"If we do it right, two or three years from now a different world will come into place," said William R. Pulleyblank, director of Exploratory Server Systems for IBM Research. "Right now, we're just happy the chips respond when you ask for their ID."

In the meantime, IBM says its efforts to capitalize on its life-science research division are paying off in other ways. At least one company, MDS Proteomics, made a joint venture with IBM research part of the package when it bought a supercomputer from the life-sciences division.

Mayo Clinic tests

During the next two months, Mayo Clinic plans to test the product of its own collaboration with IBM: a secure electronic warehouse of the medical records of every patient who has consented to be included.

Dr. Piet C. de Groen, a gastroenterologist at Mayo, had long envisioned the project as a way to accelerate the move to personalized medicine.

Then Stolovitzky, who was working with another Mayo researcher on the use of gene expression data in chronic lymphocytic leukemia, suggested that an executive in the life-sciences unit make a visit.

After that meeting, the project took on momentum, assisted by the close proximity of IBM's server division in Rochester. IBM's chairman, Samuel J. Palmisano, met with the board of governors at the world-famous clinic on a visit to its Rochester location this spring.

Mayo hopes the database, designed to include detailed genetic information, will help physicians understand how individuals are likely to respond to disease by making it easy to compare them with others of similar genetic profiles. The nonprofit Mayo is hoping the database will blend the practice and research of medicine to the benefit of both.

IBM hopes the same. It has filed for 18 patents based on the project and is already marketing its newfound expertise to businesses and governments around the world.