The Research Of 3D Protein Structure Prediction Based On Improved Artificial Bee Colony Algorithm

Protein is the material basis for the existence of life and the only form of activity, which plays an irreplaceable role in the organism. However, the protein function is determined by its conformation or spatial structure. Due to the experimental precision and technical limits, the number of proteins obtained by experimental means is still small, therefore we need to predict the protein structure by theoretical calculation or statistical forecasting methods, and provide a basis for further study of protein function and molecular design.Protein structure research technology is more difficult than DNA technology because of protein uncertainty and variability. There are two problems for protein folding prediction: the establishment of a simplified model to simulate biological protein structure and the use of reasonable optimization algorithm to solve a smaller energy value. Commonly, there are two simplified model : HP lattice and AB off-lattice model. In this paper, we choose three-dimensional AB off-lattice model to simulate protein folding process, and artificial bee colony algorithm modified to solve a low energy value, so that the predicted structure of protein can be close to the native protein. This paper makes two kinds of improvements for artificial bee colony algorithm, and the detail content is as follows:First, a hybrid algorithm is proposed which is based on the differential evolution algorithm and artificial bee colony algorithm. As the artificial bee colony algorithm does not make full use of the optimal solution in iterative process, making the convergence ability of algorithm is reduced. In order to take advantage of the optimal solution, the differential algorithm and bee colony algorithm is combined, DE-ABC algorithm is proposed. The differential evolution algorithm is used in the employment phase of the bee colony algorithm, increasing the number of variables involved in the evolution of each iteration, and accelerating the convergence rate. The original strategy is still used in the investigation phase, enhance the local search ability of the algorithm, which can fully search the solution space in the iteration, and be able to achieve a global optimal solution.Second, in order to improve the exploring ability of the algorithm, we draw lessons from the particle swarm optimization which put the optimal solution of iterative process into searching formula of bee colony algorithm, and make full use of the optimal solution in every iteration process to guide the search process. The choice of probabilistic algorithms has also been some improvements, which increase the opportunity of the lower fitness individuals is selected. Therefore, when the algorithm converges to a certain extent, it can be able to jump out of local optimal solution and increase the diversity of the population, making the algorithm can get a better optimal solution.The two improved algorithms are applied to protein structure prediction, mainly about the Fibonacci sequence and real protein sequences. The results prove the applicability of the algorithm and which provide a better way to protein structure prediction.