Protein Folding Prediction Based On Parallel Hybrid Particle Swarm Optimization Algorithm

The biological function of protein is always determined by its native conformation. It is very important for the protein function discovery and artificial protein synthesis to understand the structural features of protein. At present, protein structure prediction has become one of the hot research in related fields.Predicting protein structure by the computer simulation methods has being a research focus for the scholars. The proposed method which is based on a thermodynamics hypothesis, that is, the protein structure is the most stable when the potential energy in the protein molecule reaches the minimum value, is composed of two parts. In the first part, a 3D AB off-lattice model is selected which could reflect the natural protein conformation more accurately, and its potential energy function is constructed by the key angle force between the adjacent amino acids and the gravitation between the nonadjacent amino acids. The second part is the function optimization and the particle swarm optimization has been utilized in this paper. Considering the drawbacks of the slow convergence speed in the end stage and the undesirable solving precision, the tabu-search algorithm has been brought to improve the accuracy of solution in the last convergence stage and the particle mutation technique has been added to enhance the ability of jumping out the local minimum interval.The protein molecule usually contains a multitude number of amino acids, thus locating each amino acid in the operation process will lead to the load of the computer increase and a long calculation time. Therefore, the parallel model of the algorithm has been proposed, which using data scattered to parallel fleet in each child can effectively reduce the load of the computer processor. And through random communication mode, can expand the search area, and help improve the precision of the algorithm and the operating efficiency.Finally, the Fibonacci sequence and the real protein sequences experiments show that the proposed method could optimize the protein potential energy function more efficient than other algorithms,so according to the solution,the protein structure could be simulated and its protein conformation has meet the physical properties of proteins, which have fully proved that the method is an effective protein structure prediction methods.