Design And Implementation Of Multi-Modal Molecular Conformation Space Optimization

With the completion of Human Genome Project and the development on the field of protein research, the protein conformation determination and the relation of protein’s mechanism and its structure turn to be one of the most challenge problems in the field of molecular biology. According to Anfinsen rule, taking advantage of computer technology and optimization methods, protein structure determination already become the most active research field in bioinformation technology based on the amino acid sequence. On the aspect of mathematics view, the ab initio method in protein structure prediction belongs to the category of global optimization problems, which employs one or more optimization algorithms to search the conformation space, expecting detect the steady structure with the minimal energy. For the complex structure of protein structure, the traditional methods are not competent with this task. For these factors, it is quite important to design an efficient, high reliable global optimization based on the specific problem model.The content and the main research results of this dissertation are as follows:(1) We make a review of the background of protein structure prediction, and its development. For the tough optimization problems on high dimension conformation optimization, we introduce the current situation in molecular conformation optimization and its development in detail, and analyse the popular conformational optimization methods and it’s merit and demerit at the same time.(2) We introduce the existed molecular force field model in detail, and put focus on the ECEPP/3force field model which is employed in this paper and the transformation method from cartesian coordinates to dihedral angle coordinates.(3) We implement a classical algorithm-Conformational Space Annealing(CSA) with C++object-oriented programming language based on the Windows XP platform. CSA is quite efficient in solving protein structure determination problems. And our experiments proved this judgement and acquired the steady conformation of Met-Enkepalin via this method with ECEPP/3force field. (4) Based on the ECEPP/3force field, we introduce a hybrid optimization method CSA-DE, which takes advantage of Differential Evolution algorithm(DE) and Conformaitonal Space Annealing(CSA). With the framework of CSA, we employed DE operators in the disturbing procedure. Without lose the conformation diversity, the proposed algorithm can enhance the local convergence speed; meanwhile, Quasi-Newton method is also integrated into the methods, which is in favor of simplifying the complex optimization model. For the low dimension problems, the three benchmark functions results indicate that the proposed method can achieve good performance both in convergence speed and reliability. While in the high dimension instance, Met-Enkephalin proves the efficiency of CSA-DE.In the final, we summarize the primary research work in the whole dissertation, interpreting the research achievement and its deficiencies. And at the same time, we make a outlook of the future in the field of protein structure prediction.