Parameterizing non-canonical amino acids for cyclic peptide simulations

Non-canonical amino acids (ncAAs) are increasingly used to expand the functional chemical space available to biomolecular design. They are particularly useful for cyclic peptides, which have smaller interaction surfaces than proteins and thus require precisely tuned high-affinity contacts with very specific chemical motifs to achieve target affinity and selectivity. Despite rapid advances in machine learning for biomolecular modeling and design, physics-driven approaches remain the predominant choice for exotic chemistries because they pair generalizable, classically defined energy functions/force fields with scalable performance. High-quality, transferable parameters exist for the 20 canonical amino acids, but not for the far larger space of ncAAs; as a result, practitioners are often required to create their own residue definitions before meaningful scoring, sampling, or molecular dynamics (MD) simulations are possible. Doing so is challenging because it is a multi-step process requires careful attention to ensure compatibility of the residue definition for the target modeling tools (e.g., Rosetta and AMBER). Here we provide a guide for building custom ncAAs for biomolecular modeling and design in Rosetta and MD simulation in AMBER. Our guide is presented specifically in the context of cyclic peptide modeling. We provide sample scripts for each major step of the protocol and include validation checkpoints and representative outputs. Together, we anticipate that these resources will enable novices and experts alike to efficiently generate robust ncAA parameters for Rosetta-based modeling/design and AMBER MD simulations.