Pharmacophore model of immunocheckpoint protein PD-L1 by cosolvent molecular dynamics simulations

Due to the clinical success of cancer immunotherapy, the design of PD-1/PD-L1 inhibitors has become an area of active research. To date, only five monoclonal antibodies are approved by FDA. Despite the great effort for the development of small molecules and peptides as inhibitors, only one of those has reached clinical trials. Pharmacophore models are a proven useful tool for drug design. The effectiveness of receptor-based pharmacophore modeling is limited due to the neglect of protein flexibility and desolvation effects. In the present application, we performed co-solvent molecular dynamics simulations of PD-L1 protein in order to obtain a pharmacophore model of PD-L1 immunecheckpoint protein. The analysis of probe molecules affinities by PD-L1 resulted in the identification of C'CFG beta strands as the zone with the highest convergence of hotspots, which corresponds to PD-1/PD-L1 interaction surface. The interactions maintained with PD-L1 residues varied from hydrophobic interactions to hydrogen bonds and salt bridges with critical residues for PD-1/PD-L1 binding (M115, A121, Y123, I54, Y56, E58, R125). The superposition of known inhibitors of PD-L1 as Peptide-57, BMS-1166 and high affinity PD-1(HAPD-1) allowed us to validate the pharmacophore model due to the good correlation with its features. The pharmacophore described herein can lead to the optimization and design of more selective and potent anti-cancer drugs.