High-throughput Screening of the Mechanical Properties of Peptide Assemblies

TL;DR

This study integrates simulations, experiments, and machine learning to rapidly assess and predict the mechanical properties of peptide assemblies, facilitating the design of peptides with superior performance for various applications.

Contribution

It introduces a high-throughput workflow combining molecular dynamics, experimental validation, and machine learning to predict peptide mechanical properties based on sequence.

Findings

Calculated Young's modulus for all di- and tripeptides.

Validated computational predictions with experimental data.

Predicted properties for selected pentapeptides using machine learning.

Abstract

Peptides are recognized for their varied self-assembly behaviors, forming a wide array of structures and geometries, such as spheres, fibers, and hydrogels, each presenting a unique set of material properties. The functionalities of these materials hold exceptional interest for applications in biology, medicine, photonics, nanotechnology and the food industry. In specific, the ability to exploit peptides as viable and sustainable mechanical materials requires sequence design that enables superior performance, notably a high Young's modulus. As the peptide sequence space is vast, however, even a slight increase in sequence length leads to an exponential increase in the number of potential peptide sequences to be characterized. Here, we combine coarse-grained molecular dynamics simulations, atomic force microscopy experiments and machine learning models to correlate the sequence length and composition with the mechanical properties of self-assembled peptides. We calculate the Young's modulus for all possible amino acid sequences of di- and tripeptides using high-throughput coarse-grained methods, and validate these calculations through in-situ mechanical characterization. For pentapeptides, we select and calculate properties for a subset of sequences to train a machine learning model, which allows us to predict the modulus for other sequences. The combined workflow not only identifies promising peptide candidates with exceptional mechanical performances, but also extends current understanding of the sequence-to-function relationships for peptide materials, for specific applications.