A Study Of Optimal Model And Parallel Algorithm Of Protein Folding Problem

The thermodynamic hypothesis points out that the tertiary structure of a protein is encoded by its primary structure, and the unique tertiary structure of a protein is the conformation with minimal free energy. So it is very important how to predict the tertiary configuration by the sequence of amino acids and how to find the minimal free energy. The entire process of protein folding, which is rather random and complex, includes the course of thermodynamics and the course of dynamics. Therefore, a simple model and a feasible method are both necessary in the process of study about protein folding. The main results in the thesis are as follows:1. Traditional HP models on different lattice are all serial models, so, this thesis presents the parallel HP lattice model, which using the idea of parallel evolvement of cellular automaton. Parallel HP lattice model can more really simulate vital phenomena than serial model, and can cover some shortages of serial lattice HP model.2. The ideal of energy function should concisely express the relation between space position of protein atom and protein energy. But it is very difficult to solute the model for the complexity of protein structure, which constructed by character of protein in terms of the biology, physics and chemistry. So, the energy function should be predigested. This thesis presents a simple energy function based on the tradition energy function and the HP lattice model. This function can be solved easily by mathematic method.3. Domain decomposition algorithm has an advantage over the others when the domain of function is irregular and the function is multidimensional. So, this algorithm is used to find the minimum value of protein energy in this thesis. That IS to say, the Surface of the energy function is divided into some sub-blocks, and the minimal value is searched in these sub-blocks concurrently, then exchanges the massage each other. If the minimal value just drops into the boundary of this area, forecast the boundary value is quite necessary. In this thesis, some properties of pre-condition iteration and some new methods are introduced to construct the pre-condition matrix. For example, the localorthogonalization algorithm, approximate LU decomposition and so on. And the research also finds that parameter can simplify the certification process in discussion of condition number. In addition, the discussion of parameter's bound is also presented, and a virtual boundary forecast algorithm is used to forecast the boundary. At the same time, the validity of this algorithm is proved.