Global Optimization of Large Molecular Systems Using Rigid-Body Chain Stochastic Surface Walking

The global potential energy surface (PES) search of large molecular systems remains a significant challenge in chemistry due to “the curse of dimensionality”. To address this, here we develop a rigid-body chain method in the framework of a stochastic surface walking (SSW) global optimization method, termed rigid-body chain SSW (RC-SSW). Based on the angle–axis representation for a single rigid body, our algorithm realizes the cooperative motion of connected rigid bodies and achieves the coupling between rigid-body chain movement and lattice variation in the generalized coordinate. By exploiting the numerical energy second derivative information on rigid bodies, RC-SSW can optimize the global PES of large molecular systems with an unprecedentedly high efficiency. We show that RC-SSW is more than 10 times faster in locating the model protein global minimum while revealing many more low energy conformations than molecular dynamics and can identify low energy phases of molecular crystals up to 172 atoms missed in the sixth CCDC blind test.大型分子体系的全局势能面（potential energy surface, PES）搜索始终是化学领域的重大挑战，其核心难点在于「维度灾难」。为解决该问题，本文在随机表面行走（stochastic surface walking, SSW）全局优化方法的框架下，提出了一种刚性链方法，命名为刚性链SSW（rigid-body chain SSW, RC-SSW）。基于单个刚性体的角轴表示法，本算法实现了连接刚性体的协同运动，并在广义坐标中完成了刚性链运动与晶格变化的耦合。通过利用刚性体上的数值能量二阶导数信息，RC-SSW能够以前所未有的高效率优化大型分子体系的全局势能面。研究表明，RC-SSW在定位模型蛋白质全局极小值时的速度较分子动力学快10倍以上，且可发掘更多低能构象；同时，其可识别出第六次剑桥晶体学数据中心（Cambridge Crystallographic Data Centre, CCDC）盲测中遗漏的、包含多达172个原子的分子晶体低能相。