Fig. 2From: Clustering of disulfide-rich peptides provides scaffolds for hit discovery by phage display: application to interleukin-23Determination of clustering cutoffs. Top row: example of resulting clusters following the initial native overlap hierarchical clustering step. Each image represents a different cutoff applied for determining the final clusters for that step. These images informed the decision of which cutoff to impose in the final protocol. a Conotoxin and small hairpin clusters at the native overlap cutoff of 0.7, which was ultimately selected as the final cutoff. b At a cutoff of 0.6, the same conotoxin and small hairpin DRPs were assigned to the same cluster despite assuming different secondary structures. c At a cutoff of 0.8, conotoxin DRPs were assigned to separate clusters despite each cluster fold consisting of circular loops and short helical regions. Bottom row: example of a resulting cluster following the knottin reclustering step, with each image representing the knottin cluster containing the most DRPs after applying a different cutoff. Only the disulfide bonds in the DRPs are displayed, in yellow. The cutoffs assessed were (d) 2.0Â Ã… RMSD, (e) 1.5Â Ã… RMSD, and (f) 2.5Â Ã… RMSD. 2.0Â Ã… was selected as the optimal cutoff and used in the final protocolBack to article page