The long-term objective of the proposed research is to provide computer-based tools to aid in the high-throughput screening of drug candidates, derived from combinatorial chemistry and similar techniques. The data for the QSAR models will be derived from human liver-based chromosome P450 enzymes. Three classes of models will be developed: 1. QSAR models for the estimation of Vmax and Km of various transformations. The Phase I application deals only with hydroxylation. 2. QSAR models for the estimation of probability of a compound being a substrate (or inhibitor) for a particular isozyme. The Phase I application deals only with CYP3A4. 3. QSAR models for intrinsic clearance, for various transformations. Again, the Phase I application deals only with hydroxylation. The data for the kinetic models will be derived from the literature; the data for the substrate probability models from a previously assembled database of enzyme-compound relationships. Statistical techniques will include stepwise and all-possible regressions, the kNN-QSAR program based on a nearest-neighbor algorithm, as well as well as regression based on adaptive splines.
