Simulated Force Spectroscopy of a Badly-Behaved Protein: on the Reliability of Protein Coarse-Grained Models

Abstract

We used single-molecule pulling molecular dynamic simulations to capture the detailed structural mechanism of the unfolding process for a protein linked to neurodegenerative disease. In these simulations, a loop-truncated variant of superoxide dismutase 1 (SOD1) was pulled from the N- and C-Termini. Force-Extension Curves (FECS) and the predicted unfolding pathways of the protein were obtained for all-atom models in explicit solvent, and the results compared to several Coarse-Grained (CG) models in common use; including implicit solvent models, all-ATOM- 3-BEAD-, and Cα-based-Gō models, martini model, and associative memory hamiltonian models. In systematically comparing dynamic unfolding processes for all-atom and coarse-grained protein models, we had to resolve several problems including what energy and time means within a coarse-grained system. We describe our solution to this problem, and then discuss which cg models, if any, are most successful in predicting realistic force-induced unfolding pathways.