Assessment of the Sampling Performance of Multiple-Copy Dynamics versus a Unique Trajectory

The goal of the present study was to ascertain the differential performance of a long molecular dynamics trajectory versus several shorter ones starting from different points in the phase space and covering the same sampling time. For this purpose, we selected the 16-mer peptide Bak₁₆BH3 as a model for study and carried out several samplings in explicit solvent. These samplings included an 8 μs trajectory (sampling S1); two 4 μs trajectories (sampling S2); four 2 μs trajectories (sampling S3); eight 1 μs trajectories (sampling S4); 16 0.5 μs trajectories (sampling S5), and 80 0.1 μs trajectories (sampling S6). Moreover, the 8 μs trajectory was further extended to 16 μs to have reference values of the diverse properties measured. The diverse samplings were compared qualitatively and quantitatively. Among the former, we carried out a comparison of the conformational profiles of the peptide using cluster analysis. Moreover, we also gained insight into the interchange among these structures along the sampling process. Among the latter, we computed the number of new conformational patterns sampled with time using strings defined from the conformations attained by each of the residues in the peptide. We also compared the locations and depths of the obtained minima on the free energy surface using principal component analysis. Finally, we also compared the helical profiles per residue at the end of the sampling process. The results suggest that a few short molecular dynamics trajectories may provide better sampling than one unique trajectory. Moreover, this procedure can also be advantageous to avoid getting trapped in a local minimum. However, caution should be exercised since short trajectories need to be long enough to overcome local barriers surrounding the starting point and the required sampling time depends on the number of degrees of freedom of the system under study. An effective way to gain insight into the minimum MD trajectory length is to monitor the convergence of different structural features, as shown in the present work.