Methods for Calculating the Free Energy of Proteins

描述(申请人提供)：我们建议开发有效的方法来计算生物大分子体系的绝对熵S和自由能F，这在计算机模拟中是一个极其困难的问题。S是一种有序性的度量，它定义了蛋白质表面环的柔性程度，并构成了蛋白质折叠的基本驱动力。F(而不是能量E)是正确的稳定性标准，是预测蛋白质天然结构所必需的；它在合理的药物设计过程中是不可或缺的，例如将配体灵活对接到酶的活性部位、蛋白质-蛋白质识别过程和酶反应。计算S-lnP需要知道抽样概率F的值，这不是由蒙特卡罗或分子动力学模拟直接提供的，因此F=E-TS也是未知的。常用的方法是基于热力学积分，其中得到了m和n态之间的自由能差值。对于具有大结构方差的m和n，这些方法变得不切实际，因为沿积分路径需要大量的模拟。然而，如果在P中已知，则即使对于非常不同的状态，也只能从两个单独的模拟中获得绝对Fm和Fn，因此deltaFm，n=Fm-Fn。PT发展了计算S和F的两种近似方法，即局域态(LS)方法和假设扫描(HS)方法，并被发现对大量系统是非常有效的。对于经历局部波动(例如，在a-螺旋附近)的体系，LS的效率明显高于HS，而HS对于无规卷曲聚合物的效率最高。这个项目的一个目标是设计出LS和HS的混合方法，并且可以针对任何链的灵活性进行最佳调整。这种混合方法特别适用于由力场和隐式溶剂化作用模拟的蛋白质，将被应用于蛋白质中的线状和环状多肽和表面环进行测试。LS还将首先逐步扩展到液态Ar，然后扩展到液态水的TIP4P模型，最后扩展到浸泡在显式水的“盒子”中的多肽，在那里将计算α螺旋态和发夹状态的相对稳定性。因此，用于分子建模和药物设计的迫切需要的有效工具将变得可用。未来，PI将把这些方法结合到一种新的治疗灵活性的方法中，该方法现在应用于具有生物学意义的表面环；它们也将用于灵活对接的新程序中。

DESCRIPTION (provided by applicant): We propose to develop efficient methods for calculating the absolute entropy S and the free energy F for biomacromolecular systems an extremely difficult problem in computer simulation. S is a measure of order, which defines; for example, the extent of flexibility of surface loops in proteins, and constitutes an essential driving force in protein folding. F (rather than the energy E) is the correct criterion of stability, required to predict the protein's native structure; it is indispensable in procedures for rational drug design such as flexible docking of ligands to active sites of enzymes, protein-protein recognition processes, and enzymatic reactions. Calculating s -ln P requires knowing the value of the sampling probability F, which is not provided directly by Monte Carlo or molecular dynamics simulations; therefore, F=E-TS is also unknown. The commonly used methods are based on thermodynamic integration where the difference in free energy, deltafm,n between states m and n is obtained. For m and n with large structural variance these methods become impractical because of the large number of simulations required along the integration path. However, if In P is known, the absolute Fm and Fn, hence deltaFm,n = Fm - Fn, can be obtained from two separate simulations only even for very different states. Two approximate methods for calculating S and F, the local states (LS) and the hypothetical scanning (HS) methods, were developed by the PT and were found to be extremely efficient for a large number of systems. LS is significantly more efficient than HS for systems undergoing local fluctuations (e.g., around an a-helix), while HS is most efficient for random coil polymers. One objective of this project is to devise methods that are hybrids of LS and HS and can optimally be tuned for any chain flexibility. The hybrid methods, which are especially suited for a protein modeled by a force field and implicit solvation, will be tested as applied to linear and cyclic peptides and surface loops in proteins. LS will also be extended gradually first to liquid argon, then to the TIP4P model of liquid water, and finally to a peptide soaked in a "box" of explicit water, where the relative stability of alpha-helical and hairpin states will be calculated. Thus, much needed efficient tools for molecular modeling and drug design will become available. In the future, these methods will be incorporated by the PI within a new methodology for treating flexibility, applied now to surface loops of biological interest; they will also be used in new procedures for flexible docking.