| Investigation of protein-protein interaction and recognition is an important problem in the field of molecular biology. Given the difficulties in experimentally determining the structures of protein complexes, the docking method to computationally predict potential binding modes is currently of great interest. The final goal of molecular docking is to construct a complex using the unbound structures of the receptor and the ligand. For the unbound docking problem, it is necessary to take into account molecular flexibility in some degree. In protein-protein docking, however, because of the large number of atoms and degrees of freedom involved, it has been very difficult to simulate the conformational changes occurring during the formation of complexes. Additionally, a search procedure may produce millions of docked structures. How to drastically eliminate the incorrect structures and retain the near-native ones as many as possible before scoring them with a refined energy function is a serious challenge topic of current researches.The main goal of a scoring function is to distinguish between the near-native structures and a large number of incorrect ones within a reasonable time. Currently, the scoring functions used in many docking algorithms are based on the thermodynamic hypothesis that the native protein-protein complex is the structure with the lowest binding free energy. Thus, an important step in this approach is to develop the binding free energy functions that are computationally feasible and yet accurate enough to discriminate the near-native conformations from the incorrect ones.In the present work three parts are included. In the first part, aiming at the unbound docking problem, we have proposed an efficient protein-protein soft docking method. In the second part, the complex type-dependent docking algorithm is developed according to the soft docking method. The last part is a study on the scoring function in protein-protein docking algorithm. The main concept of this work includes the following aspects:(1) Our soft docking method is proposed based on Wodak and Janin's protein-protein rigid docking algorithm. In our method, the "simplified protein" model is used with one sphere per residue, and the molecule is of flexibility to some extent taken into account through softening the molecular surface. Because the large conformational changes occurring upon complex formation are frequently confined to the protein surface, especially for the side chains of flexible amino acids Arg, Lys, Asp, Glu and Met, the locations of the mass centers and the values of radii of these residues are modified for modeling protein surface. This tolerates overlap in some degree between the amino acid side chains on the surfaces of the receptor and the ligand. The interface residue pairing preferences are introduced into the geometric matching .filter, and a double filtering technique is implemented to eliminate most of the unreasonable binding modes. The energy minimization is performed on the retained structures, and then these structures are evaluated with a scoring function which includes electrostatic, desolvation and van der Waals energy terms. The 26 complexes were used to test this docking algorithm and good results were obtained. The results indicate that because of the improvement in two aspects mentioned above, the successful probability of the docking prediction is increased from unbound structures. For the case of enzyme trypsin-inhibitor APPI, although the large conformational change occurs to the Arg 15 side chain of APPI upon complexformation, the native-like structure was still found and ranked first.(2) On the basis of the soft docking method, the further improvement is made in the filtering stage to develop the complex type-dependent docking algorithm. In terms of the interface features of complexes, the different filtering methods based on the geometric matching, hydrophobicity and electrostatic complementarity of interfaces are applied to different types of complexes including...
|
PDF Full Text Request