CC(=O)NCCc1c[nH]c2c(Cc3cccc(-c4ccn[nH]4)c3)cc(F)cc12
CC(=O)NCCc1c[nH]c2c(OCc3ccc4nc[nH]c4c3)cc(F)cc12
CC(=O)NCCc1c[nH]c2c(Cc3nc4ccccc4[nH]3)cc(F)cc12
CC(=O)NCCc1c[nH]c2c(CNc3ccc4c(c3)C(=O)NC4)cc(F)cc12
CC(=O)NCCc1c[nH]c2c(-c3cccc(-c4ccn[nH]4)c3)cc(F)cc12
COc1cc(Nc2cc(F)cc3c(CCNC(C)=O)c[nH]c23)cc(OC)c1
CC(=O)NCCc1c[nH]c2c(Oc3ccc4ncccc4c3)cc(F)cc12
COc1ccc(Cc2cc(F)cc3c(CCNC(C)=O)c[nH]c23)cc1OC
CC(=O)NCCc1c[nH]c2c(Oc3ccc4cn[nH]c4c3)cc(F)cc12
CC(=O)NCCc1c[nH]c2c(S(=O)(=O)Nc3ccc4c(c3)C(=O)NC4)cc(F)cc12
CC(=O)NCCc1c[nH]c2c(Nc3ccc4ncccc4c3)cc(F)cc12
CC(=O)NCCc1c[nH]c2c(NS(=O)(=O)Cc3ccc4[nH]ncc4c3)cc(F)cc12
CC(=O)NCCc1c[nH]c2c(NCc3ccccc3C#N)cc(F)cc12
CC(=O)NCCc1c[nH]c2c(Nc3ccc(NC(C)=O)cc3)cc(F)cc12
CC(=O)NCCc1c[nH]c2c(CNc3cc4ccccc4[nH]3)cc(F)cc12
CC(=O)NCCc1c[nH]c2c(NCc3nc4ccccc4o3)cc(F)cc12
CC(=O)NCCc1c[nH]c2c(N(C)Cc3ccc4[nH]ncc4c3)cc(F)cc12
CC(=O)NCCc1c[nH]c2c(Nc3nc4ccccc4o3)cc(F)cc12
CC(=O)NCCc1c[nH]c2c(N(C)c3cc4c(cc3Cl)OCO4)cc(F)cc12
CC(=O)NCCc1c[nH]c2c(NCc3ccc4nc[nH]c4c3)cc(F)cc12
CC(=O)NCCc1c[nH]c2c(NCc3ncccn3)cc(F)cc12
Top 21 results hand-selected from a virtual fragment expansion, docking, and screening exercise. First, a set of 2.1K existing viral protease inhibitors from ChEMBL was analyzed to extract a library of 9.4K common design motifs. Reverie’s in-house fragment expansion engine was used to combinatorially place these fragments on the 1- and 7-indole positions of fragment X0104 via a variety of simple, synthetically-accessible linkers. Results were filtered extensively for toxicity and reactivity using an in-house library of SMARTS patterns, and an RDKit cLogP cutoff of 5.0 was applied. The remaining 18.6K compounds were docked via multiple docking methods and pose-constrained to match the orientation of the crystal-bound X0104 fragment. An ensemble of scoring functions was used to select a set of 211 compounds with high predicted affinity. These compounds were manually triaged by individuals on the Reverie chemistry team to select the final set of compounds for submission. The final set includes compounds that make favorable hydrophobic interactions with Leu27, Thr35, His41, Met49, and Gly143.
These molecules have good predicted properties to infer oral dosing, are free from structural alerts that may slow development, and are synthetically accessible from commercial building blocks so as to facilitate rapid analog synthesis, hypothesis testing and SAR follow up. Contributors: *Gabriel Grand, *Elana Simon, *Michael Bower, Bruce Clapham, Jonah Kallenbach (* = equal contribution)