Protein Sequence Optimization with a Polarizable Force Field: Insights from PDZ Domains

Abstract

Computational protein design (CPD) success is currently limited by a number of approximations, which often results in lists of candidate sequences with only a handful of genuine hits. Using a recent extension of the dead-end elimination (DEE) algorithm to many-body potentials, we are able to complement pairwise CPD energy functions with the polarizable AMOEBA force field in a self-consistent reaction field generalized Kirkwood continuum solvent. Results from energetic decompositions and preliminary sequence optimizations suggest solvent effects greatly dampen all but a handful of short-range three-residue interactions. Decompositions with a wild-type dielectric boundary only slightly reduce this damping, suggesting minimal impact of long-range 3-residue energies in implicit solvent models. In addition to these decompositions, we show preliminary sequence optimizations of PDZ domain-ligand complexes using both AMOEBA and OPLS-AA force fields. Differences are observed for both side-chain rotamer preferences and for optimal amino acid sequence. These results pose a series of questions not just about fixed-charge, pairwise decomposable sequence design models, but also our novel polarizable, many-body optimization algorithms and how to use both approaches synergistically.