Enhanced Conformational Sampling Using Replica Exchange With Collective-Variable Tempering

Recently we have found an error in the demux.pl analysis tool of Gromacs 4.6.7. This tool is used to obtain the trajectories across the replica ladder which we used to compute the distributions of the root-mean-square deviation (RMSD) showed in Figure S6 of the Supporting Information of ref (1). More precisely, the processed trajectories are not reliable if the simulations exceed 100 ns per replica. We recomputed the distributions with a corrected version of the script. A significant change in the level of agreement of the trajectories can be appreciated when comparing Figure S6 of ref (1) with the new Figure ​Figure11 where the histogram of the RMSD deviation from a representative structure has been computed with the properly processed trajectories. Three replica exchange molecular dynamics (REMD) protocols are compared in the figure, namely, temperature REMD (T-REMD), dihedral-scaling REMD (Hdih-REMD), and the newly introduced replica exchange with collective-variable tempering (RECT) method. In general, for the three compared methods the agreement between the trajectories across the replica ladder is much lower than in Figure S6 of ref (1). This indicates that all of the simulations are still not fully converged. However, when looking at the comparison between the methods, this change affects mostly our previous assessment of the Hdih-REMD method. Here the agreement decreases significantly and the method behaves thus in a way similar to that of T-REMD. Trajectories from RECT still show a better agreement among them than trajectories from T-REMD.