CUDA Based Parallel Research For Protein Structure Prediction Of AB Off-lattice Model

Protein folding problem which is also called as protein structure problem (PSP) is the hot s- pot of the combination of biology and computer science. It computes the min energy of protein to predict the structure of protein theoretically by using computer, which has stupendous sense to information biology. Because of the complexity of protein and various algorithms used for looking for the min energy on PSP, the calculation tend to be more excessive and the time spend it on tend to be longer. In the improvement of algorithms and parallel platform, OpenMP parallel programming and grid computing and cluster computing has been applied on research. However, the improvement of parallel speed still needs further promotion on cost and results. CUDA platform is a new general-program paralleling computation platform based on GPU.It brings a new direction on the parallel research of protein structure prediction problem. Because CUDA platform uses the multi-cores of GPU to do parallel computation and its programming method that similar to C-language is helpful for developer to study.In this paper, the main job is to study the brand new CUDA platform and transplant the research of PSP onto CUDA. By implementing the immune clonal selection algorithm of PSP on CUDA, we do the further exploration and study of PSP. At the same time, we have a considera- tion of paralleling of general program on CUDA, which provides a reference for the general program research on CUDA..