DESCRIPTION (provided by applicant): The need for developing rigorous QSAR modeling protocols, reliable models and accurate activity/property predictors has never been more important. This need is illustrated by two major examples of recent public efforts spearheaded both by the US and European community in the areas of bioactivity prediction (PubChem) and environmental safety (OECD Programme), respectively. To address this critical need, the overarching goal of our research is to develop a universally applicable and robust predictive QSAR modeling framework that will afford highly significant, externally validated, and predictive QSAR models of important biological endpoints. The critical components of such framework have been developed in the course of many years of our research on QSAR methodology development and application to experimental datasets. Building upon our previous experience, this proposal focuses on the design of optimized QSAR protocols for the development of reliable models (or predictors) of multiple target datasets that are useful for the virtual screening and accurate prediction of the target activities or properties for large databases or virtual libraries of untested chemical entities. These highly intertwined objectives will be achieved via concurrent development of novel QSAR methodologies (Specific Aim 1), application of these methodologies to multiple available datasets of biologically active compounds, especially of complex nature, to develop validated and predictive target-specific models, or predictors (Specific Aim 2), and virtual functional annotation of existing chemical databases (Specific Aim 3). As has always been characteristic of our research, we intend to make all our algorithms, predictors, and annotated compound databases publicly available via the C-ChemBench [ceccr.unc.edu] system that is being developed in our group with the support from the previous cycle as well as with additional funding (Center planning grant) provided by the NIH RoadMap program. We expect that the implementation of this project will advance the field of chemical genomics by developing the highly robust and publicly available predictive QSAR modeling framework, multiple validated models of diverse biologically significant endpoints, and multiple candidate compound hit lists prioritized for biological testing against the selected endpoints. PUBLIC HEALTHE RELEVANCE The need for developing robust Quantitative Structure Activity (QSAR) methodologies is illustrated by two major examples of recent public efforts spearheaded both by the US and European community in the areas of bioactivity prediction (PubChem) and environmental safety (OECD Programme), respectively. To address this critical need, the overarching goal of our research is to develop a universally applicable and robust predictive QSAR modeling framework that will afford highly significant, externally validated, and predictive QSAR models of important biological endpoints.