Representation of the conformational ensemble of peptides in coarse grained simulations

In their native state, many proteins/peptides display an ensemble of conformations, rather than a unique tertiary structure. Novel experimental techniques have enabled a quantitative analysis of this structural heterogeneity. In molecular dynamics simulations, however, capturing this conformational ensemble quantitatively remains a major challenge even with all atom simulations. In coarse grained (CG) simulations, with fewer degrees of freedom, representation of the conformational ensemble becomes more problematic. Here, we revisit a CG model from our group, which was designed to address the conformational transferability problem by using the LKα14 peptide as a model system. The LKα14 peptide transitions from a random/unstructured state in dilute solution to a solely α-helical conformation upon aggregation as evidenced by circular dichroism. Here, we demonstrate that the structure/physics based approach, used in the original parameterization of our CG model, strongly depends on the reference system chosen and excluded volume interactions that are often considered to be of secondary importance. We first tune the excluded volume parameters by using both α-helix and β-sheet type structures as reference and then update the nonbonded interactions by using a goodness-of-fit metric for representation of the conformational ensemble of LKα14. We demonstrate that the updated model can recover the whole conformational ensemble quantitatively while maintaining the aggregation driven conformational transition. This balanced parametrization with regard to alternative secondary structures opens the door for the generalization of the CG model to other sequences, which we demonstrate on a β-sheet forming triblock peptide.