Structural Bioinformatics.

We are developing computational methods to calculate the shapes of proteins and other biological molecules.  Our approach uses cutting edge developments from computational geometry and computational topology.  Studies of protein shapes have two focuses: 1) the surface regions,  including pockets, binding sites, and their precise cast or mold. The goal is to predict protein-ligand binding, protein-protein interactions and uncover novel biochemical functions based on full characterization of protein whole surfaces of the universe of all known protein structures.  2) the interior packing of proteins, and its relationship with protein stability and folding, We are also developing empirical statistical potential useful for protein fold recognition problem and for protein design.

Cheminformatics and Drug Discovery.

We apply an integrated approach for drug discovery. On the small molecule side, novel shape and chemistry based descripters have been developed to provide the metrics for managing chemical diversity of compound database and combinatorial libraries. On the receptor side, pocket surface analysis and precise cast of binding site provide additional rich information for rapid virtual screening of compounds to achieve enhanced enrichment of useful lead compounds.  Our approach emphasizes the physicochemical properties of the molecules rather than bond connectivities, and we are developing methodology that allows lead hopping where compounds of related biological activity but different underlying medicinal chemistry can be identified. Existing close collaboration with pharmaceutical industry is an important component of research in this area.

Data Mining.

We are applying various statistical pattern recognition techniques and mathematical and statistical methods for classification and prediction problems arising from high dimensional data in drug discovery.  These include discriminant analysis, parametric and nonparametric methods, hybrid models, neural nets and analysis tools complementing Principal Component Analysis and other Gaussian-distribution based methods.

Computational Biology.

We study the molecular electrostatics and solvation problem using continuum model. We are developing a boundary element method for the Poisson-Boltzman equation, with emphasis on accurate shape representation, as well as the application of fast multilevel method.   Of particular interest is the differential treatment of surface and core region of proteins embedded in solution. In addition, we are studying interactions between cosolvent and proteins in terms of both osmotic stress and preferential exclusion related hydration changes. Our approach uses detailed geometric analysis and is applied to the study of enzyme reaction to understand the relationship between water transfer and enzyme mechanism.