Global Optimization Approaches for Molecular and Protein Conformation Problems

分子和蛋白质构象问题的全局优化方法

• 批准号: 9808210
• 负责人: Panos Pardalos
• 金额: $ 10.93万
• 依托单位: University of Florida
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1999
• 资助国家: 美国
• 起止时间: 1999-10-01 至 2001-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9808210&HistoricalAwards=false
• 关键词: Global; Optimization; Approaches; Molecular; Protein

Systematic conformational search based on global optimization algorithms and heuristics is a powerful tool in modeling of biomolecules. Global optimization provides efficient mechanisms for finding global energy minima as well as new approaches for solving distance geometry problems arising in molecular conformation and molecular similarity and dissimilarity.This project will focus on the design, analysis, and implementation of efficient algorithms for computing the global optima for a class of multi-external optimization problems in molecular and protein conformation, specifically:Global optimization of nonconvex energy functions given the many degrees of freedom available to protein structures. Global optimization algorithms and heuristics for the distance geometry problem in molecular conformation.Global optimization approaches for solving molecular similarity and dissimilarity problems.Hierarchical optimization in computational biology problems.The "global approach" to treating these optimization problems is in contrast to other traditional "local approaches", for example algorithms based on gradient descent. The solutions obtained by the latter approaches are at most locally optimal. However it is well established that the above problems have an exponential number of local solutions different from the global.

基于全局优化算法和遗传算法的系统构象搜索是生物分子建模的有力工具。全局优化提供了寻找全局能量最小值的有效机制，以及解决分子构象和分子相似性和不相似性中出现的距离几何问题的新方法。本项目将专注于设计、分析和实现计算分子和蛋白质构象中一类多外部优化问题的全局最优解的有效算法，具体如下：蛋白质结构具有多个自由度的非凸能量函数的全局优化。分子构象中距离几何问题的全局优化算法和启发式算法。解决分子相似性和不相似性问题的全局优化方法。计算生物学问题中的分层优化。处理这些优化问题的“全局方法”与其他传统的“局部方法”相反，例如基于梯度下降的算法。后一种方法得到的解至多是局部最优的。然而，它是公认的，上述问题有一个指数数量的局部解决方案不同的全球。