Sampling Assessment for Molecular Simulations Using Conformational Entropy Calculations.

The extent and significance of conformational sampling is a major factor determining the reliability of long-scale molecular simulations of large and flexible biomolecules. Although several methods have been proposed to quantify the effective sample size of molecular simulations by transforming root mean squared distances between pairs of configurations into statistical/probabilistic quantities, there is still no standard technique for measuring the size of sampling. In this work, we study conformational entropy (Sconform) as a purely informational and probabilistic measure of sampling that does not require the adoption of any clustering protocol or distance metric between configurations. In addition Sconform, which is calculated from the probability mass functions associated with discretized dihedral angles, offers other potential advantages for sampling assessment (e.g., global character, thermodynamic significance, relationship with internal degrees of freedom, etc.). The utility of Sconform in sampling assessment is illustrated by carrying out test calculations on configurations produced by two extended molecular dynamics simulations, namely, a 2.0 μs trajectory of a highly flexible 17-residue peptide and the trajectory data set of the 1.0 ms bovine pancreatic trypsin inhibitor simulation provided by the D. E. Shaw research group.