Information Discovery at Aventis

Computerworld - At Aventis Pharmaceuticals Inc., IBM's DiscoveryLink database middleware goes against relational databases containing biological data, chemical structures and genomic information; stores of unstructured documents dealing with clinical trials and federal regulations; in-house expertise culled from e-mail by KnowledgeMail from Palo Alto, Calif.-based Tacit Knowledge Systems Inc.; and collaboration information from the eRoom "digital workplace" product from Cambridge, Mass.-based ERoom Technology Inc.
"We can bring all this stuff together in a very integrated framework, so that all the information that a scientist or clinician or regulatory person could potentially need is easily accessible and navigable from a simple and intuitive interface," says Peter Loupos, vice president for drug innovation and approval information systems at the Bridgewater, N.J.-based company. "Before, people went from database to database to database to get little pieces of the entire picture, then put it together manually."
Loupos says Aventis looked at several products before settling on DiscoveryLink. IBM got the nod because its information retrieval product was easy to use -- "truly a black box," he says -- and because it has advanced query optimization capabilities.
Aventis has embraced IBM's notion of "federated databases," which means leaving data in its existing places and formats, rather than bringing it into a central warehouse -- the approach that Oracle prefers. "My feeling is that when you look at the growth of information in the scientific world and the regulatory world, a data warehouse model is not going to be sustainable," Loupos says.
Tapping into data in its original source ensures timeliness, Loupos adds. "If you don't ask the right question at the right time to the right database, it's possible that right after you ask the question something else hits the database that's relevant."
And it aids maintainability, he says. "The beauty of the federated approach is it's extremely modular. We are regularly adding in new information sources, and when a new data source becomes relevant, you can just hook it in."
Loupos says getting users to agree on standards is vital in a complex data integration effort. "DiscoveryLink is important," he says, "but what really got it to work was getting our scientists, clinicians and regulatory people worldwide to sit down and think about process and standardization. That was the key success factor for integration of federated data."