Cuendet Michel A, Tuckerman Mark E
Department of Chemistry, New York University , New York, New York 10003, United States.
Swiss Institute of Bioinformatics, UNIL Sorge , 1015 Lausanne, Switzerland.
J Chem Theory Comput. 2014 Aug 12;10(8):2975-86. doi: 10.1021/ct500012b.
In molecular dynamics simulations, most enhanced sampling methods are traditionally associated with one particular estimator to calculate the free energy surface (FES), such as the histogram, the mean force, or the bias potential. Here, we start from the realization that four enhanced sampling methods, metadynamics and well-tempered metadynamics (in their extended Lagrangian form), as well as driven adiabatic free energy dynamics (dAFED) and unified free energy dynamics (UFED), can be used in combination with any of the three above-mentioned FES estimators. We compare the convergence properties of these estimators on the alanine dipeptide and a sodium ion solvation shell. We find that the mean force estimator is superior in all cases. We also show that it can be marginally beneficial to combine information from the histogram and the force, provided that both are of comparable accuracy.
在分子动力学模拟中，传统上大多数增强采样方法都与一种特定的估计器相关联，用于计算自由能面（FES），例如直方图、平均力或偏差势。在此，我们从这样一个认识出发：四种增强采样方法，即元动力学和加权元动力学（扩展拉格朗日形式），以及驱动绝热自由能动力学（dAFED）和统一自由能动力学（UFED），可以与上述三种FES估计器中的任何一种结合使用。我们比较了这些估计器在丙氨酸二肽和钠离子溶剂化壳上的收敛特性。我们发现在所有情况下平均力估计器都是最优的。我们还表明，只要直方图和力的精度相当，将它们的信息结合起来可能会有一定益处。
