By Mark D. Uehling

April 15, 2003 | Raimond Winslow had what could be called an informatic infarction. A biomedical engineer at Johns Hopkins University, and director of its new Center for Cardiovascular Bioinformatics and Modeling, Winslow wants to study changes in hundreds of heart proteins.

But he’s interested in the proteins after they have been assembled by the cell’s genetic machinery, after they’ve been briefly deprived of blood, and after they’ve changed in the never-constant environs of the body. His goal? “One of the major applications will be in studying the genetic and genomic basis of heart failure,” Winslow says. Alas, he says, there are no software tools to help him.

Enter IBM. The company has awarded Winslow and Johns Hopkins money, hardware, and software to make a more detailed digital portrait of the smallest details of the heart. The package includes the big iron, an IBM eServer p690 equipped with 16 microprocessors and 64 gigabytes of memory; a seven-node IBM eServer xSeries Linux cluster; and a robotic Virtual Tape Server subsystem that can store 28 terabytes of data. “I didn't know data management would be such a large issue in what we do,” admits Winslow, who is also a co-founder of Physiome Sciences, a company developing computational models of human cells and organs.

Certain proteins, Winslow says, seem to protect the heart from further damage. But his quest is complicated by the data, which include numerical graphs from mass spectroscopy and visual images of protein fragments called “gels.” So he’ll be using DiscoveryLink, the company’s DB2-dependent middleware, as part of the quest to integrate a variety of data from his own lab and others. Indeed, as Winslow explains it, the integration challenge is lofty. He aims to assimilate data from several biological and medical layers: not only genetic and genomic, but also proteomic, cellular, and organ- and individual-specific.

“This grant presents an opportunity to develop quantitative data analysis and models at every one of these levels,” he says.

Image Problem
As Winslow explains the work, he says working with images will be part of the challenge. “Both the genomic and proteomic studies are studies in image analysis,” he says. First, information must be extracted from the images. “Then the problem becomes: How do we combine it with the huge amount of annotation information out there?” Annotation information about genes and proteins, of course, abounds on the Internet but is not organized consistently.

Assimilating such disparate elements, he says, will require help from some of the top basic researchers at IBM and Hopkins. “We are going to have to develop ontologies for describing the data and schemas for supporting querying and analysis of the data,” he says.

But if he can put together such tools, it’s easy to see the applications in the clinic. Winslow is hoping to build applications that would allow the computer to tap a variety of databases to find all the patients between ages 40 and 50, and to further select only the patients whose genes are “up-regulated,” or expressed more than in other cells, by a factor of two. Winslow concedes such tools may exist in the pharmaceutical industry.

“That technology never sees the light of day,” he says. If successful, his efforts could put academic scientists on a more equal footing.

Inadvertently, Winslow let it slip that he will be test-driving a new IBM application: mineLink, which queries across different applications just as the DiscoveryLink software searches across databases. The company confirmed that mineLink is under development -- there are a few teasers about it on the Web -- but declined to offer additional details.

As a researcher who is also funded by NIH, Winslow says he is obligated to post his data and plans to put his cardiac genomic data online in a few months. “Our goal is to make raw data available online and let people see for themselves,” he says.

IBM’s Roy Eades, the company’s worldwide executive for university, government, and health in the life sciences, notes IBM is investing in a number of university grants. But he also seems to acknowledge that among medical research powerhouses, Hopkins is in a class by itself. “This brings us into brand-new areas we might not have thought of on our own,” Eades says. “The best way to build the best infrastructure is to work with people who push it to the edge.”