Computational modeling of PROTAC ternary complexes as ensembles using SILCS-PROTAC

Modeling ternary complexes is a critical and challenging problem. Proteolysis targeting chimeras, or PROTACs, are an emerging class of drugs that offer the potential to develop therapeutics targeting “undruggable” proteins by stabilizing ternary complexes. This involves leveraging the physiological protein degradation mechanism based on ubiquitination through stabilization of target protein-E3 ubiquitin ligase protein-protein interactions (PPI) mediated by the PROTAC. Existing computational methods for ligand design typically are not designed for the ternary complex problem and limitations in accuracy are present because they use rigid docking or are inefficient because they use full molecular dynamics (MD) simulations. Here we present a method which uses SILCS (site identification by ligand competitive saturation) that can address the challenges of ternary complex ligand design due to the use of precomputed ensembles of 1) functional group affinity patterns, termed FragMaps, for efficient and accurate ligand docking and of 2) a collection of putative PPI dimer 3D structures as docking targets. SILCS FragMaps are generated using aqueous, multi-solute grand canonical Monte Carlo (GCMC)/MD simulations to generate the FragMaps for both the target and ligase proteins. The FragMaps are then used in SILCS-PPI to generate an ensemble of candidate PPI dimer conformations. FragMaps of each dimer are generated by merging the two separate sets of FragMaps. Then a two-step SILCS-MC docking method is employed to relax PROTAC molecules into the ensemble of dimer FragMaps, and final scoring metrics are extracted from the most favorable ternary complex in the ensemble as determined by the predicted PROTAC binding affinity. Furthermore, the scoring metrics such as free energy contributions of the two warheads and linker of each PROTAC are extracted and weighted together with two physicochemical terms, the PROTAC LogP and the number of rotatable bonds, to construct an activity score. The method is benchmarked on a diverse set of ternary crystal structures with different proteins and PROTAC molecules, showing a satisfactory degree of correlation with DC50 values in cells for a wide variety of systems. Thus, the SILCS-PROTAC method is a powerful tool to facilitate ligand optimization by predicting binding geometries and energetics of ternary complexes.