Despite the recent landmark demonstration that all-atom molecular dynamics can describe the two-state folding of small, fast folding proteins, there still lacks a general conceptual theoretical framework for rapidly and accuratly simultaneously predicting folding routes and native structures starting only from the amino acid sequence without machine learning or homology methods. Most protein structure prediction methods depend on machine learning methods and homology, but these methods are incapable of deducing the folding mechanisms. Alternatively, available schemes for predicting mechanisms assume a priori knowledge of the native structure. We determine protein structure and folding pathways from amino acid sequence using a properly formulated C�-level model, without resorting to homology or machine learning schemes. Central to our algorithm is a search strategy based on the observation that proteins fold in a sequential fashion by incrementally stabilizing native-like sub-structures or 'foldons'. This principle of sequential stabilization is implemented by iteratively applying constraints learned from prior rounds of folding simulations to guide the formation of secondary and tertiary structure. Dihedral angles are sampled in multiple rounds of Monte Carlo simulations using amino acid and neighbor dependent statistical potentials. Information extracted from earlier rounds of folding simulations are used to guide the folding in subsequent rounds with biases intended to carry the system over the major free energy barrier separating the unfolded and native states in a manner that replicates the pathway of authentic folding process. Comparable folding pathways and tertiary structures are obtained for twelve proteins recently studied using benchmark all-atom molecular dynamics simulations, with our calculations running several orders of magnitude faster.

Rapid de novo simultaneous prediction of protein structure and folding pathway

246th National Meeting (2013)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.