Innovative virtual screening of PD-L1 inhibitors: the synergy of molecular similarity, neural networks and GNINA docking.

Immune checkpoint inhibitors targeting PD-L1 are crucial in cancer research for preventing cancer cells from evading the immune system. This study developed a screening model combining ANN, molecular similarity, and GNINA 1.0 docking to target PD-L1. A database of 2044 substances was compiled from patents. For molecular similarity, the AVALON emerged as the most effective fingerprint, demonstrating an AUC-ROC of 0.963. The ANN model outperformed the Random Forest and Support Vector Classifier in cross-validation and external validation, achieving an average precision of 0.851 and an F1 score of 0.790. GNINA 1.0 was validated through redocking and retrospective control, achieving an AUC of 0.975. From 15235 DrugBank compounds, 22 candidates were shortlisted. Among which (3)-1-(4-acetylphenyl)-5-oxopyrrolidine-3-carboxylic acid emerged as the most promising.