Semi-Automated Discovery of Synthetic Polymers with Protein Features

Enzymes are biological catalysts with unrivaled complexity. In some cases, tens of millions of years of evolution have shaped their structure and function. To create enzymes with new capabilities, a design strategy that can replace evolution is needed. The goal of this project is to combine artificial intelligence, polymer science, and robotics to accelerate the design and development of synthetic enzymes. Peptides are strings of amino acids, the building blocks of enzymes. Combining polymers and peptides will create a greater variety of structures and chemistries to be evaluated than would be possible simply using peptides to create enzymes. To prepare a future workforce proficient in this approach to enzyme design, this project will create an interdisciplinary and immersive environment for graduate and undergraduate students. Students from traditionally underrepresented groups will be actively recruited via outreach and participation in training programs at Rutgers University. Enzymes are typically globular proteins with an active pocket capable of catalyzing chemical reactions with exceptional specificity. Single-chain polymer nanoparticles (SCNPs) hydrophobically collapse around catalytic elements, creating synthetic enzyme mimics. There is no current method that yields bespoke polymers that can assume protein-like structures. The goal of this project is to develop SCNPs that collapse and structure around catalytic ligands to produce globular nanomaterials that mimic the structure and function of glutathione peroxidase and carbonic anhydrase. The central hypothesis is that active machine learning through physicochemical landscapes via automation will provide an efficient process for designing catalytic SCNPs with enzymatic activity. This project will implement an iterative and closed-loop Design-Build-Test-Learn process to reveal underlying structure-function behavior that encode enzyme mimetic behavior in SCNPs. The following three objectives will be completed: 1) program automation for sequence-level synthetic control of SCNPs, 2) machine learning model training through active learning on an automated platform, and 3) in-depth analysis of structure-function relationships to reveal underlying biophysical behavior. The resulting models and data will be published open source to serve the synthetic enzyme / SCNP community as a digital resource. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.