Studies On Predicted Methods For Protein-Protein Binding Site And Complex Configuration

In computational biology, the interactions between multiple proteins and the interactions between proteins and other small molecules hide a lot of biological processes. Typical problems include the protein docking calculation between two proteins, protein binding sites prediction and the protein complex conformation prediction. Prediction method of protein-protein interaction is based on the three-dimensional structure and biochemical properties of the protein molecule. With the development of computer, three-dimensional structure of the protein model and molecular docking method gradually apply in intermolecular interaction fields.Docking method is one of the important methods in the study of molecular biology. The essence is the recognition processing by using the space matching method and energy matching method between two or more molecules. Docking method in the drug design field has a wide range of applications. Space matching method is the basis for the intermolecular interaction, and energy matching method remains stable of complex. Docking methods are divided into two classes:rigid docking and flexible docking. Rigid docking method only changes the relative position and orientation of the proteins, and time-consuming is relatively small. Flexible docking method allows free changes in molecular conformation, but it is with the large time-consuming and more suitable for accurate inspection intermolecular recognition situation.Multiple structural proteins by a rigid transformation of the molecular docking method can get a more stable complex conformation. We propose that the three-dimensional molecule structures cannot change, and we can only chang the relative position and orientation between them to predict the protein complex. We design the spatial matching method for protein binding site prediction. We have a rigid molecular docking method to predict protein three-dimensional structure of the rigid transformation. This method predicts the protein docking complex conformation with good accuracy and coverage. We design the energy matching method for protein binding site prediction. Two protein structures can generate some docking cofigurations by rigid transformation. We use the energy function to evaluate these results and give reliable result. For flexible docking method, we predict the changes of the side chains in the protein-protein interaction residues, and produce more stable binding state of the complex three-dimensional conformation.Protein binding site prediction is one of the basic problems in molecular biology. The key point of protein function is able to interact with other different types of molecules, including the combination of protein molecules. The area of a protein binding with other molecules is known as the protein binding sites, while the protein three-dimensional structure determines the shape and nature of the binding sites. Protein binding sites are divided into the binding sites between proteins and the binding sites between proteins and other small molecules.The study will be divided into a single substructure searching problem and substructure pair searching problem. There exists an algorithm for discretization of the rigid transformation with O((1/ε)5n). On this basis, a new local search algorithm is proposed for the rigid transformation with O(log(1/ε)55n).An important prediction problem is the complex configuration problem for the docking between multiple proteins. Two or more protein molecules interact with each other can form a complex conformation. Protein complexes are generally divided into two categories:structural protein complexes and functional protein complexes.Three proteins can form a complex three-dimensional conformation. We first proposed a discretization algorithm with O(q13q23q3(log(1/ε)55)2). On this basis, a new distance threshold heuristic algorithm is proposed with O(q13q23q3+(Mq12q22q3+MNq12q22q3)log(1/ε)55).The main contributions of this thesis are as follows:Ⅰ. Similar substructure searching algorithm1. We first proposed a new local search algorithm is proposed for the rigid transformation with O(log(1/ε)55n).Ⅱ. The3D complex conformation prediction2. We first proposed a discretization algorithm with O(q13q23q3(log(1/ε)55)2) for predicting complex configuration. On this basis, a new distance threshold heuristic algorithm is proposed with O(q13q23q3+(Mq12q22q3+MNq12q22q3) log(1/ε)55).Ⅲ. Molecular docking method.3. Study of the rigid molecular docking method. We design the spatial matching method and the energy matching method for protein binding site prediction. Two methods predict complex configuration with better precision and recall than other existing prediction methods.4. Design of the flexible molecular docking method. We predict the changes of the side chains in the protein-protein interaction residues. The flexible docking method can predict the configurations with better recall and precision by comparing with other method.To study the work stated in this thesis, we propose the following future directions.1. We plan to capture the distribution of these dihedral angles by using the Von Mises distribution.2. We plan to adopt a probabilistic graphical model:each residue as a node, two residues within a certain distance range as an edge. Good probabilistic graphical model and the angular variation of the model can describe all the structural possibilities.3. In the protein-protein interaction network, the symmetrical complexes can be divided into Cn and Dn. We can find a circle or a double circle for predicting protein complex.4. Clustering algorithm will be improved and applied in the complex prediction:division method and density-based local search algorithm.5. Protein complex topology graph is combined with the biological properties and sequence information to predict the complex.