Explaining drugs to a computer, and helping that computer explain them to its friends: Early experience with scalable knowledge base development and implications for the Semantic Web

Mark S. Tuttle, FACMI, VP Apelon, Inc.
with John Carter, PhD('02) Apelon, Inc.
Kevin Keck, Keck Labs
Steven Brown, MD Veterans Administration / Government Computer-based Patient Record



Computers need to talk to one another about drugs. Experience suggests that both computers will need access to a reference model - AKA an ontology - if each is to "understand" drug, i.e., medication, messages sent between them. Such a reference model, more specifically a medications Reference Terminology (mRT), "defines" medications for a computer. The definition specifies "identity," what a medication is, and "function(s)," what a medication is used for. However, the critical parts of the definition specify "similarity" along various dimensions, e.g., IS_A hierarchies for chemical structure class, mechanism of action, pharmacokinetics, and therapeutic use. A preliminary version of mRT - 600 active ingredients and 1,100 "orderables" represented in Description Logic - has been deployed in a terminology server in support of demonstration applications. The next phase of the project will be to scale mRT to cover all important active ingredients and tens of thousands of "orderables," and to accommodate "new drug" transactions. The latter clarifies an important shortfall of current work on the Semantic Web. Since mRT will never be "done" and since it will be "perfection seeking" an important requirement is that it be recognized by humans and by applications as a dynamic resource. Specifically, databases of patient descriptions should not be rendered obsolete by changes to mRT. We believe mRT provides a working context in which to test and deploy Semantic Web constructs and standards.


Mark studied engineering, applied mathematics and computer science at Dartmouth College, the Thayer School of Engineering, and Harvard. He taught computer science at UC Berkeley, and medical information science at UC San Francisco. At UCSF Mark led one of four national teams working on the NLM (National Library of Medicine) UMLS (Unified Medical Language) initiative. The UCSF team led development of the UMLS Metathesaurus. In 1988, the Metathesaurus team moved to Lexical Technology. In 1999 Lexical merged with Ontyx to form Apelon. Currently, Mark is part of teams creating and applying formal terminologies for medications, mouse models of disease, and anatomical structure.