Optimization Algorithms For Drug Molecule Docking And Applications On Cloud Platform

Drug molecular design, as a new method of drug research, has been achieved a lot of research findings. Along with the development of related theories, drug molecular design is making progress and development. The rapidly development of computer technology has promoted drug molecular design method greatly. By using of computer-aided drug design, the cost of drug research has been reduced largely. Molecular docking is simulating the combination of ligand and receptor using computer technology. It is a very important method for computer-aided drug design.Cloud computing is the development and integration of grid computing, parallel computing, distributed computing and other related computer technologies. By using of cloud computing, end-users could enjoy their interested software and services on the internet. The progress is simple and freely. CUDA (Compute Unified Device Architecture) is a universal architecture of parallel computing. It help the low-cost GPU (Graphic Processing Unit) to accomplish some complex science computing tasks. Due to many advantages, cloud commuting and CUDA become to research hotspots.The main contributions of this dissertation are including:(1) Based on residue groups and evaluate model of knowledge, a new molecule docking method was proposed. As the process of simulating the drug molecule docking, using the movement of residue groups to simulate the flexibility changes of receptor. Optimization model of knowledge is used in the method to evaluate the binding affinity of molecule docking. Information entropy genetic algorithm was used to solve this optimization problem. The results of experiments show that this method has good docking accuracy and time consumed.(2) Based on some important docking factors, a self-adaption molecule docking method was proposed. Docking models of force field, experience and knowledge are used synthetically (self-adaption) to evaluate molecule docking. The results of experiments show that this method has very good docking accuracy. It can be used in some special circumstances, like refined molecule docking.(3) Under the architecture of CUDA, parallel algorithms of information entropy genetic algorithm and generating algorithm of receptor’s score grid are developed. Genetic operators, penalty function and contracting factor of information entropy genetic algorithm are analyzed and parallelized. Considering the characteristic GPU and finding the parallel components, a parallel algorithm of generating algorithm of receptors score grid is proposed. The results of experiments show that parallel algorithms have very good computing efficiency.(4) A service oriented Computational Biology Community Cloud (CBCC) is constructed. Under the consideration of its end users and resources, a four-layer cloud architecture is proposed to achieve CBCC system. Related research achievements of our group and hardware are integrated. These resources are converted to easily and useful services on the internet. Users can began or continue their computational biology works on CBCC, like molecule docking and virtual screen etc. It is easy to use, increasing users working efficiency.