Mapping the population of protein conformational energy sub-states from NMR dipolar couplings

We present a general method that exploits experimental RDCs (Residual dipolar couplings) to map the free-energy landscape occupied by folded proteins in solution, determining populations of accessible conformational sub-states contributing to the dynamic equilibrium. The method initially exploits multi-level AMD simulation, flooding the conformational space available to the protein as completely as possible to sample different sub-states, which are combined to provide an extensive pool of conformers, comprising both high-and low-energy conformations. Boltzmann-weighted ensembles are then assembled by comparison with experimental NMR data. Ensemble selection is achieved using model-free interpretation of RDCs combined with a specifically designed genetic algorithm. The approach is termed SUPERNOVA (sub-state populations on potential-energy surfaces using restraints from NMR spectroscopy and conformational over-sampling). The accuracy of SUPERNOVA, and its robustness against bias in the pools of structures from which the ensembles are selected, were tested using a synthetic dataset simulated from a hypothetical system