Robustness and generalization of structure-based models for protein folding and function

Functional dynamics of native proteins share the energy landscape that guides folding into the native state. Folding simulations of structure-based protein models, using an minimally frustrated energy landscape dominated by native interactions, can describe the geometrical aspects of the folding mechanism. Technical limitations imposed by the fixed shape of conventional contact potentials are a key obstacle toward advanced applications of structure-based models like allostery or ligand binding, which require multiple stable conformations. Generalizations of existing models, commonly using Lennard-Jones-like potentials, lead to inevitable clashes between their repulsive branches. To resolve these challenges, a new contact potential is developed that combines an attractive part based on Gaussians with a separate repulsive term allowing flexibility for adjustments of the potential shape. With this new model multiple minima for studies of functional transitions can be introduced easily and consistently. A sensitivity analysis for five small proteins confirms the robust behavior of structure-based models with our adaptable potential and explores their capacity for quantitative adjustment of the folding thermodynamics. We demonstrate its ability to incorporate alternative contact distances in simulations of structural transitions for the well-studied ROP dimer. Individual contact pairs can switch between distinct states to match the competing syn and anti structures. The flexibility of the new potential facilitates advanced uses of structure-based models. Depending on the application, features can be chosen from physical considerations or to match experiments. Generalized models can be built from multiple structures to study structural transitions or effects of disorder.