#3 · Structural bio & protein design

Simulation methods extend to cyclic peptides, non-canonical residues and polypeptoids

The common thread is extending molecular simulation and structure prediction beyond canonical linear biopolymers, toward cyclic peptides, non-canonical residues, polypeptoids, and intrinsically disordered proteins. Cyclome combines 930 curated cyclic peptides with a cyclic alignment algorithm, replica-exchange MD thermal stability runs, and an ESMc-based model, STop2Melt. A MARTINI 3 compatible coarse-grained forcefield for polypeptoids covering 19 residue types is parameterized from metadynamics-sampled all-atom data. An all-atom diffusion model, AGDIFF, retrained on CREMP data, predicts small cyclic peptides containing non-canonical residues. In the amyloid-beta study, T-REMD ensembles feed self-docking, while a sequence-based amyloid pipeline for DPF3 is followed by all-atom MD. DyME applies high-throughput MD-based mutagenesis to protein recognition. What is new is the pairing of curated datasets, generative or diffusion models, and coarse-grained parameterization with conventional sampling methods.