AI Accelerated Chemical Screening Integrates ChemBERTa to Identify Repositionable CASP4 Inhibitors via MD Simulations and MM/PBSA Analysis

Abstract This study employed an integrated ligand-based virtual screening pipeline to identify potential CASP4 inhibitors from the DrugBank database, leveraging docking-score prioritization, SMILES-derived ChemBERTa embeddings, and key physicochemical descriptors. Building on this foundation, the workflow incorporated virtual screening, cheminformatics modeling, PK–PD evaluation, molecular docking, molecular dynamics simulations, and MM/PBSA free-energy analysis to systematically prioritize repurposed DrugBank compounds for CASP4-targeted Alzheimer’s disease therapy. A Random Forest classifier trained on the hybrid ChemBERTa physicochemical feature set distinguished active from inactive compounds with ∼95% accuracy (ROC–AUC = 0.73) and achieved a ∼3.5-fold enrichment of active compounds among the top-ranked hits, while a companion Random Forest regressor, trained on the experimental pIC50 values of active compounds, ranked candidates by predicted potency. In addition, integrated cheminformatics modeling, PK–PD analysis, and molecular docking further narrowed the selection to the top five candidate compounds: DB00519, DB01068, DB06202, DB08882, and DB05316. Finally, MD simulations and MM/PBSA calculations established clear thermodynamic support for DB05316 and DB00519, whose binding free energies (−21.4 and −20.9 kcal/mol, respectively) exceeded even the reference compound donepezil, highlighting them as the most promising CASP4 inhibitors. Overall, this workflow provides an efficient and robust strategy for prioritizing repositioned DrugBank compounds as potential CASP4 inhibitors against Alzheimer’s disease.