Computational validation of top designs using REMD simulations.

For each simulation, the average RMSD is computed for backbone atoms using uncorrelated configurations sampled at temperature state 300 K. As a negative control, average RMSDs are also computed for randomly permuted sequences of each size. By comparing the free energy surfaces of the designed structures with those starting from alternative conformations, the validation results are categorized into computationally validated (the design state is reached in REMD simulations from different starting states), computationally suggestive (the design state is preserved in REMD simulations if started in the design state), and failed (the design state is not preserved in REMD simulations).