Cyclic peptides (CPs) have been shown to be promising modulators of protein–protein interactions, with high target specificity for protein surfaces. N-methylation is one way to further enhance the pharmacokinetic properties of CPs, improving both their membrane permeability and oral bioavailability. The ability to computationally predict the structural ensembles of N-methylated CPs will greatly enhance the rational design capabilities of this class of molecules. Peptide force fields, however, are typically parameterized for the trans isomer and not parameterized to provide an accurate description of the cis-trans isomer ratio. Due to ring strain, cis peptide bonds are frequently observed when CPs are N-methylated. Therefore, there is no guarantee that the current peptide force fields can accurately describe the cis-trans isomer ratio and reliably predict structures of N-methylated CPs. We tested the performance of two simulation force fields, RSFF1 and RSFF2, on predicting structures of two benchmark N-methylated hexapeptides. One hexapeptide contains all-trans peptide bonds, while the other contains one cis peptide bond. To enhance the sampling of the CPs, we performed bias-exchange metadynamics simulations, using 2D biases along coupled dihedral angles (φi × ψi) or (ψi × φi+1). When the simulations were started with the correct CP isomer from NMR structure, unexpected cis/trans isomerization was observed for the N- methylated residues. We found that cis/trans isomerization was reduced if we instead used 2D biases along modified φ and ψ angles, which better coincide with the improper dihedral potentials used to maintain the amide planarity. With this modification, we found that when the correct isomer was given for each peptide bond, RSFF2 was able to recapitulate both experimental structures, while RSFF1 was only able to recapitulate the all-trans structure. To allow the N-methylated amino acids to easily undergo cis/trans isomerization, we included additional biases along the ω angles of those residues. With this additional bias, both RSFF1 and RSFF2 inaccurately predicted the CP with the cis peptide bond to adopt the trans conformation, while the all-trans hexapeptide structure was correctly predicted. These results suggest that the cis isomer in one N-methylated cyclic peptide benchmark is overly penalized and re- parameterization of the peptide force fields is needed to reliably predict cis-trans isomers for N-methylated cyclic peptides.

Predicting N-methylated cyclic peptide structures to inform force field development

253rd National Meeting (2017)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.