| Computational prediction of macromolecular interactions is both feasible and useful, especially when it provides detailed, specific hypotheses that can be subjected to experimental validation. This thesis has presents three projects, docking the peptide VPg to the poliovirus polymerase, modeling lateral associations of polymerase fibers to form a two dimensional lattice and modeling lateral contacts in Tubulin and FTSZ. VPg docking involved algorithms that accommodate flexible ligands while polymerase-polymerase modeling required clustering and constraint propagation based on a repeatable unit-cell substructure. Applying the computational method to FTSZ and Tubulin required generalization of the method to allow any protein structure as input.; The experimental validation of the computational approach described in this thesis argues that lateral contacts can be successfully modeled using surface correlation docking followed by clustering based on geometric constraints such as parsimony and repeatability. The VPg binding and activity studies in chapter 2, the site-directed mutagenesis of the polivirus polymerase described in chapter 3 and the validation of lateral contacts in tubulin discussed in chapter 4 all present strong evidence that computational modeling is a powerful tool for hypothesis generation. Used together with site-directed mutagenesis, modeling prunes the hypothesis space and reduces the time and reagents necessary to experimentally probe a protein-protein interaction. |
