Validation of De Novo Designs of Solid-Binding Peptides

Abstract

Solid-binding peptides (SBPs) are versatile molecules that can control a range of atomic-scale interfacial processes, but they remain challenging to discover. Current approaches for discovery rely on directed evolution, which samples only a small fraction of possible sequences. Data-driven methods for therapeutic peptides are also not applicable as they rely on crystal structures whereas peptides adopt varied conformations at solid surfaces. To address this challenge, we recently combined biophysical modeling and machine learning to design plastic-binding peptides that were predicted to have strong adsorption enthalpies. Here, we evaluate these designs using steered molecular dynamics and single-molecule force measurements and identify de novo designed peptides that bind strongly to polyethylene. Experimental adhesion forces exceed those previously reported for SBPs, and adsorption free energies from metadynamics simulations support strong binding. Analysis of the designed peptides reveals blocks of non-polar and charged residues, which enables them to adopt conformations that segregate non-polar amino acids to the plastic surfaces (generating hydrophobic interactions) and charged amino acids away from the surfaces. The contact patterns within the non-polar blocks depend on sequence context and polyolefin type. Overall, we validate a general approach for de novo SBP discovery that has broad scientific and engineering applications.