New empirical descriptors to replace ab initio calculations will be generated to enable development of improved Ultra High Throughput (UHT, >200,000 compounds/hour) in silico models for Absorption, Distribution, Metabolism, Excretion and Toxicity (ADMET) properties of molecules from their structures. Partial atomic charges, atomic and molecular reactivities, pKa, and related descriptors will be generated using computer models trained from a set of about 500 molecules for which most of the ab initio geometries have been optimized. This approach has been demonstrated by a preliminary model for partial atomic charges, which improved existing models for pKa and several ADMET properties. Phase I will complete the development of the charge model, implement atomic and molecular reactivity descriptors, and develop improved models for pKa as well as for a number of ADMET properties. Phase II will expand this work by developing additional models for ADMET properties using in-house and publicly available data sources. The models and new descriptors will be embodied in a commercially available software program already used by industry, government, and academic organizations to provide rapid and accurate estimation of ADMET properties, to help prioritize drug candidates for synthesis and screening. This project seeks to develop empirical quantum level descriptors for characterizing molecules, and to employ them in improved computer models to predict Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET) properties of molecules from their structures. These improved models will enable industry and government agencies to rapidly identify compounds with poor ADMET properties, thereby eliminating wasted experiments, and contributing to reduced risks to test subjects, reduced failed clinical trials, and safer, less costly therapies for patients. ? ? ?
