MDSH Molecular Docking Simulation System – Local Search and scoring function Implement by CUDA platform

• Main Authors: Tsai Lin Feng, 蔡林鋒
• Other Authors: Hsu, Hong-Chun
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/hz3wh7

碩士 === 慈濟大學 === 醫學資訊學系碩士班 === 101 === This study is to implement a MDSH system (Molecular Docking Simulation based on Hadoop and CUDA) to process the molecular docking simulation of a protein and a ligand complex. The MDSH system is deployment in a computer cluster environment by integrated the Hadoop platform and graphic processing units (GPU). The algorithms using in this system include the Genetic Algorithm and the Monte-Carlo method. And the score estimating is using part of Amber force field included the van der Waals, electrostatic, and desolvation force field at search the best binding site between molecules. The processing of drug design is cost. By using high throughput filter in computer simulation, it can reduce the time and money cost efficiently. The common softwares use in molecular simulation including Autodock、DOCK、ZDOCK etc.“Autodock”,is the mostly used. It using the GA(Genetic Algorithms) to search the best binding site between molecules. A good simulation results always need large computing time. To improve the performance, parallel computing technology must be used. In our experiment, we tested 30 protein and ligand pairs by using a computer cluster system that contains 4 nodes. Comparing the results with the Autodock, the performance is improved obviously. That means, using the MDSH system in a computing cluster system can reduce the cost and time for molecular docking simulation, and increase the success rate of the drug design.