From Protein Structure to Function with Bioinformatics.pdf

9 Protein Dynamics: From Structure to Function 2399.4.3 CONCOORDCONCOORD (de Groot et al. 1997) uses a geometry-based approach to predictprotein flexibility. The three-dimensional structure of a protein is determined byvarious interactions such as covalent bonds, hydrogen bonds and non-polar interactions.Most of these interactions remain intact during functionally relevantconformational changes. This notion lies at the heart of the CONCOORD simulationmethod: based on an input structure, alternative structures are generated thatshare the large majority of interactions found in the original configuration. To thisend, in the first step of a CONCOORD simulation (Fig. 9.11) interactions in asingle input structure are analyzed and turned into geometrical constraints,mainly distance constraints with upper and lower bounds for atomic distances butalso angle constraints and information about planar and chiral groups. This geometricaldescription of the structure can be compared to a construction plan of theprotein. In the second step, starting from random atomic coordinates, the structureis iteratively rebuilt based on the predefined construction plan, commonlyseveral hundreds of times. As each run starts from random coordinates, themethod does not suffer from sampling problems like MD simulations and theresulting ensemble covers the whole conformational space that is available withinthe predefined constraints. However, the method does not provide informationabout the path between two conformational substates or about timescales andenergies (Fig. 9.12).9.4.3.1 ApplicationsCONCOORD and the newly developed extension tCONCOORD (t stands fortransition) (Seeliger et al. 2007) have been applied to diverse proteins. Adenylatekinase displays a distinct domain-closing motion upon binding to its substrate(ATP/AMP) or an inhibitor (see Fig. 9.13 left) with a C α-RMSD of 7.6 Åbetween the ligand-bound and the ligand-free conformation. Two tCONCOORDsimulations were carried out using a closed conformation (PDB 1AKE) as input.In one simulation the ligand (Ap 5A) was removed. Fig. 9.13 (right) shows theresult of a principal components analysis (PCA) applied to the experimentalstructures. The first eigenvector (x-axis) corresponds to the domain-openingmotion indicated by the arrow in Fig. 9.13 (left). Every dot in the plot representsa single structure. Red dots represent the ensemble that has been generated usingthe closed conformation of adenylate kinase without ligand as input. Green dotsrepresent the ensemble that has been generated using the ligand-bound structureas input. Whereas the simulation with inhibitor basically samples closed conformationsaround the ligand-bound state, the ligand-free simulation samples both,open and closed conformations, thereby reaching the experimentally determinedopen conformations with RMSD’s of 2.4, 2.6, and 3.1 Å for 1DVR, 1AK2, and4AKE, respectively. In structure-based drug design, often the reverse problem,