预反应态模型浅析:催化活性和近过渡态分子模拟

Assessment on the pre-reaction state of enzyme:could we understand catalytic activity with near transition-state molecular dynamic simulation?-a review

当今生物合成催化元件超进化分子理性设计的瓶颈在于有限的计算资源、研究时间与催化反应复杂势能面接近无穷无尽的计算需求之间的矛盾。然而,两个前所未有的数据集合有望拓新蛋白质工程人工智能化分子设计,其一是高通量定向进化实验带来的巨量高效突变体序列信息,其二是基于结构生物学的高阶量子力学计算所揭示的全原子飞秒精度反应机制。本文从催化基本理论、米氏复合物近进攻构象、催化循环效率控制点的角度浅析预反应态模型的基本概念和应用。预反应态模型尝试利用在低反应势垒生物化学反应中内禀的近进攻构象与过渡态具有相近的物理化学稳定性,弹性地选择与催化元件进化目标相关的关键过渡态,利用经典分子动力学模拟分析近过渡态的活性构象布居数与远端突变、底物结构、实验条件的关系。预反应态分析的基本流程为:首先,基于高阶量子力学反应势能面提取催化中心关键过渡态的结构特征;其次,从高精度蛋白质三维结构出发,结合氨基酸质子化生物信息学预测工具构建出关键过渡态对应的近进攻态活性构象;最后,利用过渡态结构特征设定分子动力学模拟初始约束条件,并逐步取消约束条件测试预反应态随氨基酸突变和底物变化的稳定性变化,以近进攻构象在预反应态轨迹中布居数作为“预反应态-酶活”半定量相关系数,从预反应态稳定性中挖掘酶与底物的适配图谱。当前在预反应态动态结构与酶活的定量关系分析上还有诸多难题亟待突破,利用高通量高阶量子化学再采样计算、结合机器学习人工智能分析代表了预反应态模型的发展方向。

The bottleneck of enzyme design for biosynthetic elements lies in the incompetence of the limited computing resources with demanding for an in-depth computation on complicated potential energy surfaces of catalytic reactions.However,two unprecedented achievements are expected to expand artificial intelligence machine learning in protein engineering-one is a variety of high-efficient mutants brought by high-throughput directed evolution experiments,and the other is the high-quality molecular simulation of all-atom with femtosecond precision revealed by ab initio quantum mechanics calculation and three-dimensional structural information.This work briefly describes the basic concept and application of the pre-reaction state(PRS)model from the perspectives of the fundamental enzyme theories,the near-attack conformation of Michealis complex,and the control points of the catalytic cycle efficiency.The pre-reaction state model tries to use the intrinsic features of biochemical reactions with low activation energy in which transition state and pre-reaction states share similar physiochemical stability,flexibly selects the rate determining transition states related to the evolutional goal of the catalytic element,and employs classical molecular dynamics simulations to understand the relationship of active conformation population with distal mutations,substrate spectrum,and experimental conditions.The general pre-reaction state protocol is:first,the near-transition state structural features are extracted from the high-level quantum-mechanical calculation on the rate-determining transition structures;then the PRS molecular dynamic simulations are collected from the restrained to the free state,which is used to study the adaptability between mutants and substrates.The population in the PRS trajectory is used as a semi quantitative correlation coefficient of“pre-reaction state-enzyme activity”(PRS-EA),and the adaptation map of enzyme and substrate is mined from the pre-reaction state stability.Although the mechanism-based pre-reaction state analysis provides an insightful rationale at atom levels as a post-NAC approach,the quantitative relationship between the PRS structure and enzymatic reaction cannot be fully illustrated owing to the ambiguity of the PRS constraint,the repeatability of molecular dynamics simulation,and the arbitrariness of reactive population.The high throughput quantum calculation for transition state samplings and machine learning and artificial intelligence could be integrated to unveil the quantitative structure-activity relationship,paving a way for the practical applications of pre-reaction state in protein engineering.