“…By identifying a network of states and transition rates, including encounter complexes and non-native binding pathways, it is possible to compare the effects of mutants on multiple interaction pathways (i.e., non-native binding vs. native binding, or the rate of transitions between different native-like bound states), as compared to utilizing simplifying assumptions of single on-rates, off-rates, folding rates, and unfolding rates when comparing simulations and experiments. In particular, we believe that MSMs derived with the methods presented here, when combined with the recently developed augmented Markov model formalism, where MSM state populations and transition rates refit using maximum-entropy methods to match agreement with experimental NMR spin relaxation data, will provide an eloquent approach to assess the agreement between MD simulations and NMR ( 118 , 119 ). To illustrate this, we have predicted backbone NMR chemical shifts for each state identified by our deep MSM ( SI Appendix , Fig.…”