Conformation-dependent environments in folding proteins

Abstract

We introduce a semiempirical approach to ab initio prediction of expeditious pathways and native backbone geometries of proteins folding under in vitro renaturation conditions. The algorithm incorporates a discretized codification of local steric hindrances which constrain the movements of the peptide backbone. Thus, torsional motion is shown to be conditioned by the hopping from one basin of attraction (R-basin) to another in the Ramachandran map or local potential energy surface associated with each residue. Rather than simulating detailed dynamics, we simulate the time evolution of such torsional constraints. The semiempirical potential needed to obtain geometric realizations of such “modulo R-basin” topologies is rescaled with each iteration of the simulation in order to incorporate the role of conformation-dependent local environments. Thus, the extent of local desolvation within which a specific interaction occurs is computed for each iteration using an effective local “solvophobic field” determined by two-body interactions emerging from the previous iteration. The predictive power of the algorithm is established by (a) computing ab initio folding pathways for mammalian ubiquitin that yield a stable structural pattern reproducing all of its native features in spite of some adverse local propensities associated with those features when taken in isolation; (b) determining the nucleating event that triggers the hydrophobic collapse of the chain; and (c) comparing coarse predictions of stable folds of moderate size proteins (N∼100) with structures from the Protein Data Bank.