Biomolecular Structure-Function Relationships and Structure Based Function Prediction
Apr 29, Tue 2008
11:00 am, MRB 200 Conference Room
Dr. Alexander Tropsha
Professor and Chair, The Laboratory for Molecular Modeling, Division for Medicinal Chemistry & Natural Products, School of Pharmacy, University of North Carolina at Chapel Hill,
Elucidating robust and predictive relationships between molecular structure and its biological function is one of the central problems of biomolecular informatics. Frequently, global structural (dis)similarity is insufficient to draw rigorous conclusions about associated (dis)similarity of biological function. Our approach to the analysis of structure-function relationships relies on searching for function-defining substructural features (or in general, selected descriptors) using a combination of datamining and statistical validation techniques. A specific general approach that I shall discuss in detail uses frequent subgraph mining approaches applied to families of biomolecules represented by labelled undirected graphs. I shall talk about the applications of this strategy in three formally different areas, i.e., predictive Quantitative Structure Activity Relationship (QSAR) modelling of organic molecules, protein-ligand interface analysis and ligand pose scoring, and structure based functional annotation of proteins.