We are developing new analytical and experimental strategies to aid in drug discovery for cancer and AIDS. The analytical approaches include both artificial intelligence and classical statistics. Initial results: l. Neural networks able to predict mechanism of drug action on the basis of patterns of activity in DTP's 60-cell line cancer drug screen. 2. A program package (DISCOVERY) that integrates information on the chemical structure, activity, and molecular targets of compounds tested by NCI. It currently has 6 modules and, as the name suggests, was designed to search for novel mechanisms of action among the 35,000 compounds tested to date. 3. Clustering in combination with neural nets and discriminant analysis to identify candidate cell lines for replacement in the screen by breast, prostate, target-selected, and target-transfected lines. 4. Neural nets and statistical methods that predict the clinical activity of phase II-evaluable drugs on the basis of patterns of activity in the screen. 5. Quantitative structure activity relationship (QSAR) studies and pharmacophoric searches in the NCI's DIS database to identify new inhibitors of the HIV-l integrase, an enzyme essential to the life cycle of the virus. With resect to experimental strategies we are: l. Characterizing molecular targets in cell of the NCI cancer and AIDS screens (by analysis of mRNA, protein, and phenotype). A major aspect is 2-D gel electrophoresis to develop a large-scale resource on hundreds of proteins in the cells. 2. Developing a hollow fiber model of solid tumors for use in vitro and in vivo testing of new therapies. These """"""""information intensive"""""""" strategies are being used to create an information interface between the cancer and AIDS drug discovery programs. Supported in part by the NIH Intramural AIDS Targeted Antiviral Program.
