Figure 1. Stability analysis of dipeptides within CBD3 and pharmacophore model on average cluster center ofA1R2 dipeptide and sample of matched hits. (A) Largest cluster of MD snapshots with RMSD less than 1 Å for dipeptides A1R2, R2S3, S3R4 and R4L5 , also shown is average percentage of cluster size in 3 independent simulations. (B) RMSD relative to cluster center and times the amino acid side chains are free of contacts (3.8 Å for hydrogen bond donors and 4 Å for hydrophobic side chains) within 3 ns windows in a representative MD trajectory of CBD3 peptide. Gray regions highlight regions for which RMSD of snapshots are less than 1 Å from cluster center, and side chains are 80% or more free of contacts within 3 ns windows. (C) Pharmacophores radiuses used: guanidine group, 0.78 Å for positive ion and 1 Å for hydrogen bond donor; 0.78 Å for other hydrogen bond donors, and 1.0 Å for hydrophobic atoms. (D) Sample of compounds that matched all pharmacophores using ZincPharmer.
TheA1R2 structural motif shown inFigure 1A , defined as the conformation having the largest 1 Å radius cluster of snapshots in our MDS, encompassed about 44% of our total runs. Furthermore, since previously reported cell-based experiments also involved the cell-penetrating tat sequence conjugated to CRMP2 derived peptides, we repeated the MDS on the full tat-ARSRLA sequence, obtaining the same structural motif as for CBD3 (Figure S1 ). The robustness of theA1R2 motif led us to use it as the basis for the design of the pharmacophore model shown inFigure 1C . We entered this design in the open access server ZincPharmer (http://zincpharmer.csb.pitt.edu/) to search for suitable compounds among more than 27 million commercially available compounds from the ZINC database (Sterling & Irwin, 2015), obtaining ~200 hits. Based on availability and manual curation, we selected 77 compounds for experimental validation (Figure 1D ). Figure S2 shows the full list of compounds.