Novack Dylan, Raddi Robert M, Zhang Si, Hurley Matthew F D, Voelz Vincent A
Department of Chemistry, Temple University, Philadelphia, Pennsylvania 19122, United States.
J Chem Inf Model. 2025 Jun 23;65(12):6089-6101. doi: 10.1021/acs.jcim.5c00704. Epub 2025 Jun 3.
Alchemical free energy calculations are essential to modern structure-based drug design. Such calculations are usually performed at a series of discrete intermediates along a nonphysical thermodynamic pathway to estimate the free energy difference between two end points of an alchemical transformation. The efficiency and accuracy of the free energy estimate depends critically on the choice of alchemical intermediates. In this paper, we review the concept of thermodynamic length, and how it can be used as a principle to choose alchemical paths in free energy simulations. We then present an algorithm for optimizing the choice of alchemical intermediates in free energy simulations. Our method is similar to the thermodynamic trailblazing algorithm of Rizzi et al. (2020), but with several improvements for use with expanded ensemble (EE) simulations. Our method only requires a single initial round of EE simulation and includes a method for optimizing the number of alchemical intermediates in an EE simulation based on the predicted mixing time. We first show how the method performs in a simple toy model, and then demonstrate its use in a realistic example for an alchemical relative thermostability free energy calculation. We also show how our method can be used to optimize free energy estimates in other contexts, namely, calculating a score for model selection in the Bayesian Inference of Conformational Populations (BICePs) approach. We have implemented our optimization algorithm in a freely available Python package called (https://github.com/vvoelz/pylambdaopt).
炼金术自由能计算对于现代基于结构的药物设计至关重要。此类计算通常沿着一条非物理的热力学途径在一系列离散中间体上进行，以估计炼金术转化两个端点之间的自由能差。自由能估计的效率和准确性关键取决于炼金术中间体的选择。在本文中，我们回顾了热力学长度的概念，以及它如何作为一种原则用于在自由能模拟中选择炼金术路径。然后我们提出了一种在自由能模拟中优化炼金术中间体选择的算法。我们的方法类似于Rizzi等人（2020年）的热力学开拓算法，但有一些改进以用于扩展系综（EE）模拟。我们的方法只需要一轮初始的EE模拟，并包括一种基于预测混合时间在EE模拟中优化炼金术中间体数量的方法。我们首先展示该方法在一个简单玩具模型中的表现，然后在一个用于炼金术相对热稳定性自由能计算的实际例子中演示其用途。我们还展示了我们的方法如何用于在其他情况下优化自由能估计，即计算构象群体贝叶斯推断（BICePs）方法中模型选择的分数。我们已将我们的优化算法实现为一个名为 （https://github.com/vvoelz/pylambdaopt）的免费Python包。
