【作者】 赵铮；

【导师】 刘海燕；

【作者基本信息】 中国科学技术大学 ， 生物信息学， 2008， 博士

【摘要】 酶是生物体内高效的催化剂,它能显著地催化各种生化反应过程,促进生物体的新陈代谢。在大多数的生命活动中,酶促反应都起着至关重要的作用。透彻的理解和领悟酶的“工作”机制是进一步理解生命过程的根本性的要求;也是进一步应用酶的高效、专一性服务于生命活动过程的内在要求。实验上,三维酶结构的解析技术(例如:x-ray,NMR等)和基因工程都取得了巨大的进展。不过,酶的结构与其催化功能是如何相关的,以及在一个具体的催化反应里不同的催化机理的可行性的如何甄别,这些都不是完全清楚,计算机模拟为我们开辟了除实验、理论之外的又一新的途径。酶是至少包括了数千上万个原子的大分子而且是非均质的,它的非凡的催化能力不仅由活性中心决定而且也受蛋白和溶剂控制。这就要求计算中要考虑整个体系。QM/MM方法折中而有效的考虑了计算精度的需求与花费的可行性,已成为研究溶液及酶体系反应的一个重要手段。在QM/MM方法计算中,多个理论水平层次的QM方法从半经验水平的方法到从头算的分子轨道或密度泛函理论都能被应用。不过一直以来,准确的从头算水平的QM/MM研究由于其需较多的机时花费大多被用于单点计算和势能面的决定,在计算自由面中也只是用优化QM部分、采样MM部分的策略,从头算水平的QM与MM部分的同步采样也局限于相对比较小QM部分的一维的自由能面的获得。半经验水平的QM/MM的计算机时花费是较低的,能有效的被用于自由能面的采样,不过半经验方法的准确度是体系依赖的,在不同的体系中它的结果可能是不可靠的。这就要求应用多理论层次的QM/MM如何有效而准确来获得势能面和自由能面。我们在这个方面进行了一系列的工作,包括了多理论层次的QM/MM研究酶的催化机理。在第二章中,我们详述这个工作。我们用多理论层次的QM/MM方法研究了L-丝氨酸脱水酶(SDH,EC 4.3.1.17)的催化机理。L-丝氨酸脱水酶(SDH,EC 4.3.1.17)是一种重要的5′-磷酸吡哆醛(PLP)依赖的酶。在这个酶的催化反应研究中,我们获得了重要的发现,包括在Schiff碱的形成步骤中底物丝氨酸的羧基的化学角色和在β-羟基消去步骤中一个涉及磷酸基团的质子依赖的机理。后面这步反应是随着α-氢的消去而后自发的(无能垒的)进行的,这也说明这是一个连续进行的α,β-消去反应。我们也获得了这个催化过程的限速步骤和被酶环境稳定的过渡态。在这个过渡态,底物丝氨酸的羧基和Lys41的氨基的电荷被局域化,使得它们与酶环境特定的相互作用能够显著的降低活化能。另一方面,在从头算QM/MM方法研究复杂的酶反应中一个主要的困难是准确合适的反应坐标的选择,在这个研究中,我们利用了半经验AM1/MM和路径优化技术克服了这个困难。我们的工作也涉及了多理论层次的QM/MM方法计算多维的自由能面来获得复杂体系(例如酶体系)的热力学性质。自由能的计算是研究热力学性质的首要任务,同样是分子模拟领域最困难的方向之一。在第三章里,叙述了我们提出的应用自适应伞形采样方法来探索多维的自由能面用以研究溶液中或酶中的复杂的化学反应问题。自适应伞形采样是一种高效的采样稀少事件的重要方法。我们为了计算其准确的自由能面,我们巧妙地设计了一个逐步递增阀来严格的控制在研究体系中的自适应伞形势增加。在采样过程中,我们应用了AM1/MM这个半经验模型,为了避免了AM1模型本身的缺点,我们加入了特异的反应参数,使得不失构象采样的高效性的同时也保证了采样势的正确。为了进一步提高效率,一个全局背景势也被应用,它作为反应坐标的函数一定程度上弥补了自由能面上深的局部极小、从而降低了能垒。我们用甲酸二聚体(FAD)测试了我们的方法,得到了收敛的自由能面,溶剂效应被分析,获得了与已报道的一致的结果。我们期望这个方案能被广泛的用于溶液中或酶中的复杂反应。更多还原

【Abstract】 Enzyme is an efficient catalysis in living organisms. In most biological systems, enzyme is all very important because it can catalyze most biochemical processes and accelerate the metabolism of living organisms. Intensively understanding the activity of enzymatic mechanism is not only a challenge but a fundamental requirement for quantitatively understanding biochemical processes and further applying enzymatic efficiency, specificities. Experimentally, Great progresses have been achieved both in the analytic techniques for protein structures (such as x-ray, NMR etc) and in the field of gene engineering. However, the correlations between the enzymatic structure and its catalytic functions are not fully clear, and in another aspect, how to validate the right mechanism in a specific catalytic process is also a challenge. In past few decades, computer simulation is playing an increasing role to study enzymatic mechanisms, which compensate the limitation of experimental and theoretical measures.Enzymes are very large and heterogeneous, containing at least thousands of atoms. Its remarkable capability is not only determined by its active site, but also affected by its protein and solvent environment. Therefore, it requires the computational method to take the heterogeneous enzyme environment into account explicitly. Conbined quantum mechanical/molecular mechanical (QM/MM) mechods have become the method of the choice of modelling of reactions in enzyme. In QM/MM simulations of reactions, multiple levels of theory approaches, from semiempirical level to ab initop molecule orbitals and density functional methods, have been used. Ab initio QM/MM is more accurate and relable; however, its computational cost is expensive so that most ab initio QM/MM studies have been limited to determine potential energy surfaces rather than FESs. Calculation of free energy surfaces using ab initio QM/MM are also limited to treat the QM parts by minimizations and the MM parts by sampling. Simultaneous sampling of QM and MM parts has been limited to systems with relatively small QM centers and for which one-dimensional FESs can be envisioned. The computional cost of semiempirical level methods is low, and semiempirical QM/MM is more efficient to calculate FESs and allows simultaneous sampling. It is, however, well-known that the accuracies of semi-empirical models are system-dependent, and results may become unreliable because of large errors. So it is very important to efficiently apply multiple levels of theory approaches to obtain the right potential energy surfaces and free energy surfaces. We have made many efforts to the field including multiple-levels QM/MM studies of enzyme mechanics. In Chapter 2, we detailtedly describes the work for studing enzymatic mechanism. In the work, the catalytic mechanism of a pyridoxal 5’-phosphate-dependent enzyme, L-serine dehydratase, has been investigated using ab initio quantum mechanical/molecular mechanical (QM/MM) methods. New insights into the chemical steps have been obtained, including the chemical role of the substrate carboxyl group in the Schiff base formation step, and a proton-relaying mechanism involving the phosphate of the cofactor in theβ-hydroxyl-leaving step. The latter step is of no barrier and follows sequentially after the elimination of theα-proton, leading to a single but sequentiala,β-elimination step. The rate-limiting transition state is specifically stabilized by the enzyme environment. At this transition state, charges are localized on the substrate carboxyl group, as well as on the amino group of Lys41. Specific interactions of the enzyme environment with these groups are able to lower the activation barrier significantly. One major difficulty associated with studies of complicated enzymatic reactions using ab initio QM/MM models is the appropriate choices of reaction coordinates. In this study, we have made use of efficient semi-empirical QM/MM and pathway optimization techniques to overcome this difficulty.Our effort also involved the methodological development for free energy calculation based on multiple-levels QM/MM. Free energy calculation is the first assignment to achieve the thermodynamical prosperities and is also one of the most difficult fields in computer simulation. In our work, we presented an efficient scheme to calculate multidimensional free energy surface as described in Chapter 3, in which we detailedly described a refined adaptive-umbrella-sampling method presented to explore multidimensional free energy surface for complex chemical reactions in solution and in enzyme. Adaptive umbrella sampling method is an efficient, important tool to sample rare events. To our target, we skillfully designed a so-called refined adaptive umbrella potential to achieve convergent free energy surface. In the scheme, a strictly adaptive umbrella potential and a global background potential are used to fit the system including chemical reaction. Moreover, to overcome the deficiencies of AM1 model, we applied the AM1 model of specific reaction parameters. The formic acid dimer is used to test the scheme. We obtained the convergent free energy surface, which is in accord with the reported results. The solvent effects are analyzed and validated. We hope the scheme can be extensively used in complex chemical reaction in enzyme.更多还原