The de-novo design protocol uses a sequential optimization approach with decision trees to explore parameter space. The design loop leverages RFDiffusion, different flavors of ProteinMPNN, and AlphaFold2.