Virtual screening (VS) based on molecular docking is one of the most useful methods in computer-aided drug design. By allowing to identify computationally putative ligands binding to the proteins of interest, VS dramatically reduces the time and expense of the development of novel therapeutics. Among the limitations of the VS approaches is the low accuracy of scoring functions implemented in docking methods for assessing binding affinity. Many such scoring functions are developed for rapid, high-throughput evaluation of binding energy of multiple conformations generated by a searching algorithm. The methods for more rigorous calculation of binding affinity calculation are generally time-consuming. Even so, in many studies more accurate methods were used for rescoring of the final poses and false-positive hits evaluation. We performed VS for three benchmark sets and used energy minimization with MM/PB(GB)SA methods (molecular mechanics energies combined with the Poisson–Boltzmann or generalized Born and surface area) to rescore binding affinities. The comparison of the area under the curve (AUC), enrichment factor (EF), and Boltzmann-enhanced discrimination of receiver operating characteristics (BEDROC) showed essential improvements in the binding energy prediction after the rescoring. Finally, we provide a program for minimization and rescoring VS results based on freely available AmberTools. The code requires just the final binding poses of the ligand as the input and can be used with any docking program.

基于分子对接的虚拟筛选（VS）是计算机辅助药物设计中最有用的方法之一。通过允许识别与目标蛋白质结合的计算推定配体，VS 显着减少了开发新疗法的时间和费用。VS 方法的局限性之一是在用于评估结合亲和力的对接方法中实施的评分函数的准确性较低。许多这样的评分函数被开发用于快速、高通量评估由搜索算法生成的多个构象的结合能。更严格的结合亲和力计算方法通常很耗时。即便如此，在许多研究中，更准确的方法被用于对最终姿势和误报命中评估进行重新评分。我们对三个基准集进行了 VS，并使用能量最小化和 MM/PB(GB)SA 方法（分子力学能量与泊松-玻尔兹曼或广义玻恩和表面积相结合）来重新评估结合亲和力。曲线下面积 (AUC)、富集因子 (EF) 和 Boltzmann 增强的接收器操作特征鉴别 (BEDROC) 的比较表明，重新评分后结合能预测有了实质性的改进。最后，我们提供了一个基于免费提供的 AmberTools 的最小化和重新评分 VS 结果的程序。该代码只需要配体的最终结合姿势作为输入，可以与任何对接程序一起使用。