Programmed cell death ligand 1 (PD-L1) is a crucial target for cancer therapy. Here, an in silico study investigates PD-L1 to inhibit its interaction with PD1, thereby promoting an immune response to eliminate cancer cells. The study employed machine learning (ML) -based QSAR to detect PDL1 inhibitors. Morgan's fingerprint with docking score showed a 0.83 correlation with the experimental IC50, enabling the screening of 3200 natural compounds. The top three compounds, considered 2819, 2821 and 3188, were selected from the ML-based QSAR and subjected to molecular docking and simulation. The binding scores for 2819, 2821 and 3188 were -7.0, -9.0 and -8.9 kcal/mol, respectively. The stability of the ligands during a 100 ns simulation was assessed using RMSD, showing that 2819 and 2821 maintained stable patterns comparable to the control inhibitor. Notably, 2819 exhibited a consistent stable pattern throughout the simulation, while 2821 showed stability in the last 40 ns. The control compound showed the highest number of hydrogen bonds with proteins, whereas compounds 2819 and 2821 formed continuous H-bonds. 3188 was separated from the protein in later phases and is not regarded as a potential PD-L1-binding molecule. MMGBSA binding free energy for complexes was computed. Control had the lowest binding free energy, while 2819 and 2821 also had lower binding energies. In contrast, 3188 showed poor binding free energy, causing protein separation. Principal component analysis showed a loss of entropy and reduced protein conformational variation. Overall, 2819 and 2821 are potential binders for PD-L1 inhibition and immune response triggering.Communicated by Ramaswamy H. Sarma.
Keywords: QSAR; RMSD; docking; in silico; machine learning; programmed cell death ligand 1.