SAMP: Structure-Adaptive Materials Prediction

Intelligent data mining tools and materials design strategies using crystallographic and materials property data sources can be enriched by developing a rational cyber-design for determining the nature and types of data equivalencies in the structural chemical information of materials. A visual portal enhancing the use of intelligently organized correlations of structural chemical data for predicting better or new materials will be applied within a Web-based learning system for educational and development purposes. This approach will be tested in several graduate and undergraduate chemical courses. One example in the area of biomaterials is the detection of key functionally active amino acids within protein structures, which precedes protein classification, evolutionary study and drug design experiments. Here, a novel approach to prediction of functional sites within proteins is presented, which involves the application of a number of machine learning algorithms to the topological space created through the Delaunay tessellation of proteins Cα backbone. To test the utility of the method, this study explores the development of methods for site identification in proteins having known structures. A large number of attributes may be extracted from the tessellation: among these the topological score provides considerable insight into active site participants and the potential for further defining the structure/function