Training Neural Network Models Using Molecular Dynamics Simulation Results to Efficiently Predict Cyclic Hexapeptide Structural Ensembles

Abstract

Cyclic peptides have emerged as a promising class of therapeutics. However, their de novo design remains challenging, and many cyclic peptide drugs are simply natural products or their derivatives. Most cyclic peptides, including the current cyclic peptide drugs, adopt multiple conformations in water. The ability to characterize cyclic peptide structural ensembles would greatly aid their rational design. In a previous pioneering study, our group demonstrated that using molecular dynamics results to train machine learning models can efficiently predict structural ensembles of cyclic pentapeptides. Using this method, which was termed StrEAMM (Structural Ensembles Achieved by Molecular Dynamics and Machine Learning), linear regression models were able to predict the structural ensembles for an independent test set with R 2 = 0.94 between the predicted populations for specific structures and the observed populations in molecular dynamics simulations for cyclic pentapeptides. An underlying assumption in these StrEAMM models is that cyclic peptide structural preferences are predominantly influenced by neighboring interactions, namely, interactions between (1,2) and (1,3) residues. Here we demonstrate that for larger cyclic peptides such as cyclic hexapeptides, linear regression models including only (1,2) and (1,3) interactions fail to produce satisfactory predictions (R 2 = 0.47); further inclusion of (1,4) interactions leads to moderate improvements (R 2 = 0.75). We show that when using convolutional neural networks and graph neural networks to incorporate complex nonlinear interaction patterns, we can achieve R 2 = 0.97 and R 2 = 0.91 for cyclic pentapeptides and hexapeptides, respectively.

Molecular Dynamics Simulations of Polyelectrolyte Complexes

“…By varying charge, hydrophobicity, and “blockiness” of a given peptide, we next plan to screen a set of strategically designed charged peptide sequences to characterize in more detail how neighboring amino acid sequences can shift single chain secondary structure. This approach of peptide screening has already been shown to be a viable method toward the rational design of self-assembled materials and the prediction of structural ensembles . Further, this method has recently gained momentum due to the combined availability of powerful computing resources and machine learning methods.…”

“…This approach of peptide screening has already been shown to be a viable method toward the rational design of self-assembled materials 46 and the prediction of structural ensembles. 47 Further, this method has recently gained momentum due to the combined availability of powerful computing resources and machine learning methods. Regular molecular dynamics could lead to the coacervate droplets being stuck in a metastable state.…”

Assessing the Performance of Peptide Force Fields for Modeling the Solution Structural Ensembles of Cyclic Peptides

“…Our previous work shows a structural digit map based on the backbone (ϕ, ψ) dihedrals can efficiently denote cyclic peptide conformations. , To construct the structural digit maps for cyclic pentapeptides, cyclic hexapeptides, and cyclic heptapeptides, we first simulated cyclic pentaglycine, cyclic hexaglycine, and cyclic heptaglycine, respectively. This choice was made due to the high flexibility and achiral nature of glycine.…”

“…According to the distributions, the (ϕ, ψ) spaces were discretized into 10, 6, and 5 regions for cyclic pentapeptides, hexapeptides, and heptapeptides. Reprinted in part with permission from ref ( 36 ). Copyright 2023 American Chemical Society.…”