Integrating Machine Learning and Structural Dynamics to Explore B-cell Lymphoma-2 Inhibitors for Chronic Lymphocytic Leukemia Therapy

Chronic lymphocytic leukemia (CLL) is a malignancy caused by the overexpression of the anti-apoptotic protein B-cell lymphoma-2 (BCL-2), making it a critical therapeutic target. This study integrates computational screening, molecular docking, and molecular dynamics to identify and validate novel BCL-2 inhibitors from the ChEMBL database. Starting with 836 BCL-2 inhibitors, we performed ADME and Lipinski's Rule of Five (RO5) filtering, clustering, maximum common substructure (MCS) analysis, and machine learning models (Random Forest, SVM, and ANN), yielding a refined set of 124 compounds. Among these, 13 compounds within the most common substructure (MCS1) cluster showed promising features and were prioritized. A docking-based re-evaluation highlighted four lead compounds-ChEMBL464268, ChEMBL480009, ChEMBL464440, and ChEMBL518858-exhibiting notable binding affinities. Although a reference molecule outperformed in docking, molecular dynamics (MD), and binding energy analyses, it failed ADME and Lipinski criteria, unlike the selected leads. Further validation through MD simulations and MM/GBSA energy calculations confirmed stable binding interactions for the leads, with ChEMBL464268 showing the highest stability and binding affinity (ΔGtotal = - 80.35 ± 11.51 kcal/mol). Free energy landscape (FEL) analysis revealed stable energy minima for these complexes, underscoring conformational stability. Despite moderate activity (pIC₅₀ values from 4.3 to 5.82), the favorable pharmacokinetic profiles of these compounds position them as promising BCL-2 inhibitor leads, with ChEMBL464268 emerging as the most promising candidate for further CLL therapeutic development.