#23 · Structural bio & protein design

In silico protein and peptide binder design with structure-based validation

The common thread is computational binder design followed by in silico validation, applied to protein and peptide scaffolds rather than to a single force field or target. A VHH nanobody against the PI/RuvC interface of SpCas9 was designed with BoltzGen and Boltz-2 and checked by 10 ns MD. A benchmark of four AI peptide-design tools (RFdiffusion, BindCraft, PepMLM, RFpeptides) scored 290 BMPR1A-binding designs with AlphaFold 3 and physics-based analysis. A GluN1-mimetic decoy peptide against anti-NMDAR autoantibodies was developed using AlphaFold2, HADDOCK docking and PRODIGY scoring. What is new is the pairing of generative design with structure prediction, docking or MD across distinct targets. Validation depth differs among the items: the GluN1 work relies on docking and PRODIGY scoring rather than molecular dynamics.