(169ab) Integrating Biophysical Modeling and Machine Learning to Discover Plastic-Binding Peptides for Microplastic Remediation

Plastics on the micrometer or nanometer scale, termed microplastics or nanoplastics (MNPs), are a harmful pollutant. Millions of tons of MNP pollution pervade the environment. Humans and many other organisms regularly consume MNPs unknowingly, which can have harmful effects on their health. Because the risks of MNP to the environment and human health worsen with increasing MNP concentration, we urgently need methods to remediate MNP pollution. Here we discover plastic-binding peptides (PBPs), which could detect and capture MNP pollution, by combining biophysical modeling and machine learning. PBPs were designed for polyethylene, polystyrene, polypropylene, and PET using the Peptide Binding Design (PepBD) algorithm. PepBD samples amino acid sequences and adsorbed conformations on a plastic surface using Metropolis Monte Carlo. Peptides were scored by the sum of their MMGBSA binding energy and peptide stability in the adsorbed conformation (i.e., the peptide internal energy). Molecular dynamics (MD) simulations of designed PBPs show that they consistently have stronger binding free energies than random amino acid sequences, while preliminary atomic force microscopy experiments found the best PBP for polyethylene has greater affinity than random sequences. An additional round of PBPs were designed to 1) increase PBP solubility in water to improve detection and capture of MNP pollution in aqueous environments, and 2) increase PBP binding affinity to a specific plastic to enable separation or characterization of the components of MNP waste. Water solubility was increased by adding the CamSol solubility term into the peptide scoring function. Binding specificity was improved by first training a long short-term memory (LSTM) network to predict the PepBD score of an amino acid sequence, and then searching for sequences with large score differences between plastics. Thorough experimental testing of PBP designs is ongoing. We envision that the designed PBPs will be broadly useful in addressing MNP pollution.