(394f) Active Learning Cycle to Design Stable Collagen-Mimetic Peptoid Triple Helices

Collagen-Mimetic Peptides (CMPs) are short synthetic analogs to full-length natural collagen with applications in engineered tissues, biofilms, and sutures. Akin to natural collagen, CMPs are characterized by a triple-helix tertiary structure and often contain a large number of N-substituted proline residues. The apparent necessity of nitrogen substitution in CMPs introduces the possibility of hyperstable collagen-mimetic peptoid triple helices. Peptoid CMPs offer the potential to engineer novel biomaterials with attractive properties such as resistance to enzymatic degradation, optimized durability, and tunable bioactivities. The peptoid sequence-structure relationship is, compared to peptides, relatively poorly understood, making it challenging to prospectively identify which peptoid sequences are likely to assemble into stable and tunable peptoid CMPs.

In this work, we employed high-throughput simulation and active learning to computationally design and experimentally test hyperstable peptoid CMPs. Our computational design loop employs simulation-based assessment of the peptoid CMP melting temperature within an active learning cycle for machine learning-guided sequence design. Experimentally, the top candidates are synthesized, and their structure and stability is assessed using circular dichroism spectroscopy. Using this workflow, we demonstrate the design and production of the first highly thermostable peptoid CMPs with a range of future biomaterials applications.