分子空间结构比较方法优化与点部署的并行实现

分子空间结构相似性比较的指标函数可以定量地描述两个分子空间结构相似性大小，但难题在于如何确定分子初始相对位置，以保证得到全局最优叠合．由于以往都是利用程序随机生成初始相对位置，所以没有规则也不确定．为保证得到全局最优叠合，需要大量的初始相对位置优化计算．利用“均匀设计”的实验设计手段，在空间规则地筛选部署分子初始相对位置，使其具有代表性和均匀分布性，这样只需少数确定数目的初始相对位置，便可以稳定地得到指标函数的全局最优解．而后利用并行处理方法，把初始相对位置集合部署在Np个处理器上同时工作，大幅度地减少了运行时间，并输出与串行执行相同的运算结果．

The Optimization for Molecular 3D-Structure Comparison Method and Its Parallel Implementation of Vectors Deployment

Molecular similarity index describes the similarity between two molecules quantitatively. But acquiring the global optimal index is a complicated problem confusing the scientists. No matter what kinds of methods are used to search the best superposition of two molecules like grid-based integral or efficient iterative technique, selecting the relative situations of molecules is needed. Because of the limitations about the random search, which has always been used, it is required to test a great lot of relative situations. In this paper, an experimental method of uniform design is applied to deploy regularly the initial relative situations of molecules in 3D space, which makes the relative situations of molecules representative and even. It ensures to get global optimization within few certain numbers of initial relative situations of molecules. Parallel implementation for deploying these situations on Np processors reduces largely the running time, which can give the same final computational result as serial program. "Uniform design" combined with parallel computation searches the best superposition between two molecules rapidly with high possibilities.