PubMed2000
This paper is an extension of the keynote address and another talk at the Symposium on the Use of Toxiciological Information in Drug Design. The symposium was organized by American Chemical Society's Chemical Information Division at the 220th National Meeting of the American Chemical Society in Washington, DC, August 20-24, 2000. We outline an approach for meeting the scientific information needs of the U.S. Food and Drug Administration (FDA). Ready access to scientific information is critical to support safety-related regulatory decisions and is especially valuable in situations where available experimental information from in vivo/in vitro studies are inadequate or unavailable. This approach also has applications for lead selection in drug discovery. A pilot electronic toxicology/safety knowledge base and computational toxicology initiative is underway in the FDA Center for Drug Evaluation and Research (CDER) that may be a prototype for an FDA knowledge base. The objectives of this effort are: (i) to strengthen and broaden the scientific basis of regulatory decisions, (ii) to provide the Agency with an electronic scientific institutional memory, (iii) to create a scientific resource for regulatory and applied research, and (iv) to establish an internal Web-based support service that can provide decision support information for regulators that will facilitate the review process and improve consistency and uniformity. An essential component of this scientific knowledge base is the creation of a comprehensive electronic inventory of CDER-regulated substances that permit identification of clusters of substances having similar chemical, pharmacological or toxicological activities, and molecular structure/substructures. Furthermore, the inventory acts as a pointer and link to other databases and critical non-clinical and clinical pharmacology/toxicology studies and reviews in FDA archives. Clusters of related substances are identified through the use of: (i) an extensive index of alternative names for each substance, (ii) a molecular structure key field consisting of a rudimentary or core structure represented as an ISIS.mol-file, (iii) global search terms (molecular group, chemical class, clinical indication, or pharmacologic activity), and (iv) molecular clustering using structure/sub-structure similarity indices. The information contained in a toxicology knowledge database has limited value unless means are available to extract information, identify relationships, and create and test hypotheses. One such means is computational toxicology, also called in silico toxicology, ComTox, or e-TOX. Computational toxicology is the application of computer technology and information processing (informatics) to analyze, model, and estimate chemical toxicity based upon structure activity relationships (SAR). A computational toxicology software package, MCASE, has been evaluated and successfully improved by CDER through the incorporation of data from FDA archives and concomitant alterations of the logic used in the interpretation of the results to reflect the data analysis and hazard identification practices and priorities of the Center. Our modifications and uses of the MCASE program are discussed in detail.