Heuristic energy-based cyclic peptide design
Fig 2
CyclicChamp computation time and design comparisons with Rosetta.
(a) The computation time required by CyclicChamp backbone sampling and stability validation (ClusterGen) exhibits linear-like growth with increasing backbone size. FastDesign was faster for 20 and 24 residues than for 15 because there were fewer backbones on which we did sequence design. (b) Total design time divided by the number of stable designs validated by the filtering method for 7 residues, ClusterGen for 15 residues, and reshaped ClusterGen for 20 and 24 residues. (c) When allocating equivalent computation time for backbone sampling, CyclicChamp generated 5 to 28 times as many cyclic backbones with sufficient H-bonds as Rosetta’s simple_cycpep_predict, which led to 2 to 11 times as many stable designs as Rosetta’s after stability validation.