# data-driven-CVs Supporting material for the paper Data driven collective variables for enhanced sampling by Bonati, Rizzi and Parrinello (2020). Contents: The zip file contains the input files and the results reported in the paper. The folder "ala2" is related to alanine dipeptide and "aldol" to the aldol reaction. For each physical system we report the following data (except for specific files): 1) Unbiased trajectories in the metastable states - Trajectory with the values of the descriptors in the two metastable states [INPUTS.*] 2) Trained Deep-LDA model - Serialized Pytorch model [model.pt], which can be used as a CV for enhanced sampling - Full checkpoint of the NN [model_with_weights.pt] to reproduce the analysis on the feature importances --aldol only-- 3) Enhanced sampling inputs - PLUMED input files [plumed.dat] - GROMACS input [ala2.tpr] --ala2 only-- - CP2K input [Aldol_pm6.inp] --aldol only-- - Additional .cpp files for PLUMED (they are loaded directly by PLUMED as instructed in the input file plumed.dat) 4) Results of the simulations - Time evolution of relevant variables [COLVAR*] - Free energy surfaces [fes*.dat] - Free energy difference over time [deltaF.dat] --ala2 only-- FILES FORMAT: All but the .pt files are text files. The .pt files can be loaded in Pytorch or LibTorch. Please visit https://github.com/luigibonati/data-driven-CVs for code updates and an interactive Python notebook for the training of the Deep-LDA collective variable.