In the field of molecular biology, there is an urgent need for tools that can help elucidate the structural basis for macromolecular binding. This project will develop computational methods for automated discovery of structural and physical-chemical elements contributing to the affinity and specificity of macromolecular binding. Graph models for the representation of protein structures and graph kernel-based machine-learning methods will add to the analysis and prediction of binding sites. The proposed graph models will provide a succinct data structure to encode a range of structural and physical properties germane to molecular interactions. Particularly, the models will reflect the flexibility of protein structures. The proposed innovative graph-kernel-based approach will investigate the modular organization of binding sites and discover characteristic patterns associated with the modules. The methods and procedures produced in this work will assist researchers in pinpointing the location of the binding sites and elucidating the binding mechanism.
Comprehensive interdisciplinary educational and outreach plans will target undergraduates, graduates, and researchers from industry. Interdisciplinary bioinformatics courses will bring together students from life sciences and computer science and spearhead the curriculum development for the bioinformatics program at North Dakota State University (NDSU).