Sound-stage Virtual Screening Based on Implicit Ligand Theory

 描述（由申请人提供）：化学文库的虚拟筛选是发现配体（如药物先导物）的常见起点。虚拟筛选的主要工具是分子对接，这是快速的，但对小分子如何与蛋白质相互作用进行了一些认真的近似。另一方面，基于严格统计物理的方法要慢得多，也更准确。我们最近推导出一种新的统计物理结合理论，该理论基于受体的多个刚性结构：隐式配体理论基于该理论的计算方法有可能在对接速度和其他严格方法的准确性之间进行仔细的折衷。在这里，我们开发和评估计算工具来预测（i）分子是否会与蛋白质结合，（ii）它结合的紧密程度，以及（iii）它结合的位置。具体来说，我们将首先测试结合到单个刚性结构上是否足以区分活性分子和非活性分子。其次，我们将评估不同的分子模拟方法产生与配体结合相关的受体结构的能力，并导致准确的结合亲和力预测。第三，我们将根据我们的理论开发排名绑定姿势的算法。我们预计我们的方法将用于虚拟屏幕的第二阶段;在对接数百万种化合物后，可以使用我们的方法处理前几千种化合物。

 DESCRIPTION (provided by applicant): Virtual screening of chemical libraries is a common starting point for discovering ligands, such as drug leads. The primary tool for virtual screening is molecular docking, which is fast but makes a number of serious approximations about how small molecules interact with proteins. On the other hand, methods based on rigorous statistical physics are much slower and more accurate. We have recently derived a new statistical physics theory for binding that is based on multiple rigid structures of a receptor: implicit ligand theory Computational methods based on this theory have the potential to carefully compromise between the speed of docking and the accuracy of other rigorous methods. Here we develop and assess computational tools to predict (i) whether a molecule will bind to a protein or not, (ii how tightly it binds, and (iii) where it binds. Specifically, we will first test whether binding o a single rigid structure is sufficient to distinguish active from inactive molecules. Second, we wil assess the ability of different molecular simulation methods to generate receptor structures relevant to ligand binding and lead to accurate binding affinity predictions. Third, we will develp algorithms to rank binding poses based on our theory. We anticipate that our methods will be used in the second stage of a virtual screen; after docking millions of compounds, one can use our methods with the top several thousand.