Calculation of Protein-ligand Binding Affinity

这项研究的目的是了解分子识别在生物学中的基础。总体目标是从参与反应的分子的结构和化学性质计算结合常数，以了解高亲和力结合的生化基础。这项研究将使用结合的统计机械处理方面的最新理论进展，生物分子计算机模拟方法的发展，以及关于蛋白质-配体复合体的最新实验数据。基本的方法是利用计算机模拟，通过对体系的所有相关自由度(平移、转动、内部和溶剂)进行积分来计算蛋白质和配体之间的适当加权相互作用自由能，从而直接得到结合常数的计算。分子力学方法将用于前三个组分。隐含的溶剂法将被用来使后者--最昂贵的集成--变得容易处理。最近的隐式溶剂化算法第一次足够准确和快速，可以在分子力学积分步骤中使用，因此整个计算可以在实际的计算机时间内以自洽的方式完成。生物识别是蛋白质和其他大分子的基本性质。所有生物都含有蛋白质，许多生命过程依赖于正确的蛋白质在正确的时间‘识别’正确的分子。一种蛋白质可以说识别另一种分子，当它比存在的许多其他分子更强烈地吸引另一种分子，并与其更紧密地结合在一起时。强大的吸引力或结合力取决于蛋白质在形状和化学性质上与其结合伙伴互补。然而，即使知道蛋白质及其靶标的详细结构，也不足以确定它们将结合得有多紧密。要确定这一点，人们需要知道吸引或排斥这两个分子的所有不同的力，并准确地总结它们。由于涉及数千个原子，这是一项艰巨的任务。方法是在计算机上构建蛋白质及其结合伙伴在其环境中的模拟，编码物理和化学定律以描述所有原子如何相互作用，然后运行计算机模拟以确定两者结合的紧密程度，对照实验测试结果，改进描述并重复，反复使模拟更真实。这个模拟蛋白质识别的软件将提供给更广泛的研究和教学社区。逼真地模拟蛋白质识别的软件将帮助研究人员在分子水平上理解活体中的基本过程，比如细胞如何交流，并帮助培训学生分子识别的原理，以及用计算机模拟分子。具体地说，这个软件将用于宾夕法尼亚大学研究生和本科生的年度课程，该课程由PI开发并讲授，标题为“分子建模”。

The goal of this research is to understand the basis of molecular recognition in biology. The overall objective is to calculate the binding constant from the structure and chemical properties of the molecules taking part in a reaction to understand the biochemical basis for high affinity binding. The research will use recent theoretical advances in the statistical mechanical treatment of binding, developments in methods for computer simulation of biomolecules, and recent experimental data on protein-ligand complexes. The basic method is to evaluate, using computer simulations, the appropriately weighted interaction free energy between protein and ligand by integrating over all the relevant degrees of freedom of the system: translational, rotational, internal and solvent, leading directly to evaluation of the binding constant. Molecular mechanics methods will be used for the first three components. Implicit solvent methods will be used to render the latter, most expensive integration, tractable. Recent algorithms for implicit solvation are, for the first time, accurate and fast enough to be used within the molecular mechanics integration step, so that the entire calculation can be done in a self-consistent way in a practical amount of computer time. Biological recognition is a fundamental property of proteins and other macromolecules. All living things contain proteins and many life processes rely on the right protein 'recognizing' the right molecule at the right time. A protein may be said to recognize another molecule when it attracts that molecule more strongly, and binds to it more tightly, than the many other molecules present. Strong attraction or binding depends on the protein being complementary in its shape and chemical properties to its binding partner. However, even if the detailed structures of the protein and its target are known, it is not enough to determine how tightly they will bind. To determine this, one would need to know all the different forces that act to attract or repel the two molecules, and sum them up accurately. Since thousands of atoms are involved, this is a formidable task. The approach is to construct a simulation of the protein and its binding partner in their environment on the computer, encoding the laws of physics and chemistry to describe how all the atoms interact, then 'run' computer simulation to determine how tightly the two bind, test the results against experiment, improve the description and repeat, iteratively making the simulation more realistic. This software for simulating protein recognition will be made available to the wider research and teaching community. Software to realistically simulate protein recognition would help researchers understand at a molecular level basic processes in living organisms, such as how cells communicate, and help train students in the principles of molecular recognition & simulation of molecules with computers. Specifically, this software will be used in the yearly course for graduates and undergraduates at the University of Pennsylvania, developed and taught by the PI entitled "Molecular modeling."