|講演者||木原 大亮 博士 (Associate Professor, Department of Biological Sciences/Computer Science, Purdue University)|
The increasing number of protein structures solved by structural genomics raises significant challenges in elucidating their functions and in effective use of the structures for modeling of structure of related proteins. In this seminar, we will discuss our new computational methods we have developed in order to meet the new challenges in the omics era. First, we will talk on our method named 3D-SURFER for real-time protein structure database search, which employs a surface representation using the 3D Zernike descriptors (3DZD). 3DZD are series expansion of a three dimensional mathematical function that can represent a 3D object in a rotation and translational invariant fashion. We will discuss cases that function of proteins are inferred by the global surface similarity and also its potential application to comparing and searching low-resolution structures of proteins which may be obtained by electron microscopy.
Next, 3DZD can be also applied for structure-function prediction by local pocket matching. We will report our recent results on comparing different representation of protein pockets, all of which allow a real-time search against a database of pocket shapes. Finally, we applied 3DZD for shape-based protein-protein docking. Our new docking algorithm, named LZerD (local 3D Zernike descriptor-based Docking algorithm), uses 3DZD for capturing local shape complementarity of interacting protein surfaces.
森 浩禎 (email@example.com)