PHYS 136

Native conformations are associated with the global minimum on a smooth free energy surface. Since this surface is prohibitively high-dimensional for exhaustive exploration, our initial efforts use a priori information in the form of an average native structure. A first-order approximation is employed to generate non-concerted fluctuations around a native structure. An efficient geometric exploration followed by energetic treatment allows to obtain large conformational ensembles. Ensembles obtained for proteins of various folds agree very well with NMR data probing native fluctuations at nanosecond-millisecond timescales.

We lift the first-order approximation in the context of cysteine-rich cyclic peptides and reduce a priori information to a topological feature like backbone cyclization. Applications to naturally-occurring and engineered sequences (18-31 aas long) show the framework correctly predicts the native state from such minimal information.

We are generalizing the framework to use only amino-acid sequence information for a protein. A hierarchical multiscale exploration strategy first obtains a broad view of the conformational space relevant for the native state. An iterative exploration then adaptively zooms in on emerging energy minima and further populates the conformational space until convergence. Our preliminary results are promising. The framework captures the main native conformations for proteins with multiple functional states.