Discovery of potential PD-L1 small molecule inhibitors as novel cancer therapeutics using machine learning-based QSAR models: A virtual drug repurposing study.

Abstract

Binding of programmed death 1 (PD-1) to programmed death ligand-1 (PD-L1) triggers downstream signaling pathways and results in inhibition of new cytotoxic T-cell activation. Cancer cells can “trick” the immune system with the T-cell suppression ability of the PD-1/PD-L1 signaling mechanism. Blocking the PD-1/PD-L1 interaction through target-specific small molecules can recruit cytotoxic T cells and reactivate the immune response, is a potential therapy in many types of cancer. In spite of that, there are currently no approved small molecule inhibitors targeting the PD-1/PD-L1 pathway. In this study, we aim to discover target-specific and effective small molecules with anti-cancer activity. With this objective, we performed a virtual drug repurposing study on PD1 protein, PD1-PDL1 structure and PDL1-PDL1 homodimer complexes. The results revealed that PDL1-dimer structure is the most suitable target model for virtual screening of small molecule inhibitors. Three different approaches with different algorithms were used in virtual screening to discover the best hit molecules: (i) structure-based, (ii) structure-based pharmacophore (hybrid) and (iii) machine learning-based QSAR. At the end of the study, which included three different approaches, promising PDL1 small molecule inhibitors such as anidulafungin, bemotrizinol, deferoxamine, dequalinium, neratinib, ubiquinol and vilanterol were discovered.