分子模拟中的变分隐式溶剂模型、算法及其应用

Implicit-solvent description is effective in modeling molecular interactions between solutes and solvent. Such models are widely used in molecular simulations due to their high computational efficiency and capability of reducing the statistical averaging errors arising from the incomplete sampling of solvent conformations. One of the most important concepts in implicit-solvent models is the solute-solvent interface, whose traditional definitions, however, cannot reflect local hydrophobic effects, electrostatic interactions and so on. To overcome these drawbacks, we propose a variational implicit-solvent model, in which the solute-solvent interface is obtained by minimizing the solvation free energy with respect to all possible solute-solvent interfaces. We develop an efficient level-set method to compute the solute-solvent interface and corresponding solvation free energies as well. In addition, we introduce solvent fluctuations to our model, and develop an explicit-solute implicit-solvent simulation algorithm. We apply the model and algorithm to study dewetting transitions in hydrophobic confinements and characterize the effect of solvent fluctuations on the binding process of ligands to receptors. Also, we propose a novel algorithm that can identify the binding pockets on protein surfaces, and assess the probability of being bound with ligands based on the local geometry and physiochemical properties inside the identified pockets. Preliminary results are in very good agreement with crystallized receptor-ligand complexes. Also, we apply our model to investigate the binding affinity of ligands to receptors as well as various contributions to the binding affinity, revealing the binding mechanisms of different types of binding pockets. We integrate all of the newly developed models and algorithms into an open-source software package to facilitate their applications in the fields of molecular solvation analysis and drug design.

隐式溶剂模型是描述溶质与溶剂相互作用的常用模型。它具有计算效率高、能降低溶剂构像采样统计误差等优点，因此在分子模拟中得到广泛的应用。溶质与溶剂间界面是隐式溶剂模型最为重要的概念之一，但传统的界面定义不能反映局部疏水效应和静电作用等影响。针对这些不足，我们提出变分隐式溶剂模型，发展水平集方法来极小化溶解自由能，从而计算得出最优界面位置和溶解自由能。在此基础上，我们引入溶剂涨落效应，发展显式溶质-隐式溶剂模拟算法，研究疏水受限区域的去湿过渡现象，分析溶剂涨落对配体与受体结合过程的影响。另外，我们提出新算法来确定蛋白质表面的结合口袋，并根据口袋的局部几何信息、物理化学性质来综合评估口袋被结合的可能性，初步结果与晶体结构高度吻合。我们计算配体与受体的结合亲和力以及各因素的贡献，揭示不同类型结合口袋的不同结合机制。我们将发展的模型与算法编写成开源软件包，供同行开展分子溶解分析和药物设计方面的研究。

隐式溶剂模拟是研究水溶液中生物分子动力学的重要工具，它具有计算效率高、能降低采样统计误差等优点。但是，传统隐式溶剂模型存在未能考虑生物分子的局部疏水效应和界面的曲率校正等问题。为了克服这些缺点，项目组在前期工作中发展了变分隐式溶剂模型和高效的数值计算方法，研究生物分子与水溶剂间的界面。在本项目中，项目组与合作者发展了动态界面隐式溶剂模型来刻画水溶剂涨落和粘性对生物分子动力学的影响。基于能量变分方法，发展了一系列连续介质模型和保物理性质数值算法来研究生物分子体系内的离子动力学和静电相互作用。另外，由于水分子在局部疏水环境中不能形成稳定的氢键网络，局部疏水区域存在多种溶解状态，并会发生去湿-浸润过渡等现象。项目组在变分隐式溶剂模型的基础上，发展了弦方法来抓取多溶解状态间的过渡现象，如两块疏水板和碳纳米管等疏水腔体内的干湿过渡。我们对多重过渡路径，能量势垒，过渡态的几何形状都作了深入地探讨。另外，我们研究疏水口袋-配体体系，探讨配体与疏水口袋的结合-分离动理学。我们提出一类多状态布朗动力学模拟算法，计算结合与分离的平均首达时间，计算结果与分子动力学模拟数据吻合。进一步研究表明：疏水口袋内的水溶剂涨落效应将减慢配体与口袋的结合过程，但会加速配体与口袋的分离过程；配体运动的非马尔可夫性源自于高维马尔可夫过程在低维空间中的投影；在配体的结合与分离过程中，配体与口袋的非平衡溶解状态具有记忆效应，呈现出磁滞现象。这些研究进展为理解水溶液中配体与口袋的结合-分离动理学提供了全新视角，对药物设计研究有重要的意义。