A Parametric Rosetta Energy Function Analysis with LK Peptides on SAM Surfaces

Abstract

Although structures have been determined for many soluble proteins and an increasing number of membrane proteins, experimental structure determination methods are limited for complexes of proteins and solid surfaces. An economical alternative or complement to experimental structure determination is molecular simulation. Rosetta is one software suite that models protein-surface interactions, but Rosetta is normally benchmarked on soluble proteins. For surface interactions, the validity of the energy function is uncertain because it is a combination of independent parameters from energy functions developed separately for solution proteins and mineral surfaces. Here, we assess the performance of the RosettaSurface algorithm and test the accuracy of its energy function by modeling the adsorption of leucine/lysine (LK)-repeat peptides on methyl- and carboxy-terminated self-assembled monolayers (SAMs). We investigated how RosettaSurface predictions for this system compare with the experimental results, which showed that on both surfaces, LK-α peptides folded into helices and LK-β peptides held extended structures. Utilizing this model system, we performed a parametric analysis of Rosetta's Talaris energy function and determined that adjusting solvation parameters offered improved predictive accuracy. Simultaneously increasing lysine carbon hydrophilicity and the hydrophobicity of the surface methyl head groups yielded computational predictions most closely matching the experimental results. De novo models still should be interpreted skeptically unless bolstered in an integrative approach with experimental data.

Macromolecular modeling and design in Rosetta: recent methods and frameworks

“…RosettaSurface 18 samples protein conformations ab initio in both the solution and adsorbed states (Figure 3D) in order to account for adsorption-induced conformational changes. Experimental data can be incorporated into the simulation 19 to improve scoring, which remains difficult because the Rosetta scorefunction has been optimized for proteins in aqueous solution.…”

“…(D) Model of an LK-α peptide (LKKLLKLLKKLLKL with a periodicity of 3.5 assuming a helical conformation) on a hydrophilic self-assembled monolayer surface. In solution, the peptide is unstructured, whereas when on the surface, experiments show that it assumes helical structure 19 . (E) Flexible docking of a carbohydrate antigen to an antibody.…”

Key aspects of the past 30 years of protein design

“…Additionally, several protocols have been developed for the interpretation of a wide range of chemical and biological macromolecular systems. This group includes the modelling of interactions with peptides [58,[68][69][70][71][72][73][74][75][76][77] and nucleic acids [78][79][80][81][82][83][84][85][86], the antibody modelling [80,[87][88][89][90][91][92][93][94] and design [32,[95][96][97][98], the modelling of membrane proteins [99][100][101][102], carbohydrates [103,104] and metalloproteins [49].…”

A Deep SE(3)-Equivariant Model for Learning Inverse Protein Folding

“…Owing to the conventional wisdom that a protein’s native state corresponds to its free energy minimum, CPD tasks are traditionally framed as an energy minimization problem. In this setting, the energy function typically consists of some combination of physics-based terms (J.H. et al 2018; Alford et al 2017; X.…”