Machine Learning-Driven Drug Repositioning Identifies Putative IRAK4 Inhibitors Through Structure-Based Computational Evaluation

Abstract

Interleukin-1 receptor-associated kinase 4 (IRAK4) is one of the IRAK family proteins and plays an important role in the regulation of innate and inflammatory responses. In particular, IRAK4 acts as a key regulator of the Toll-like receptor (TLR) and interleukin-1 receptor (IL-1R) signaling pathways and has attracted attention as a therapeutic target for immune and inflammatory diseases. In this study, an integrated computational approach combining machine learning, molecular docking, and molecular dynamics simulations was applied to identify putative IRAK4 inhibitor candidates. Bioactivity data of IRAK4 were obtained from the ChEMBL and PubChem databases and evaluated for multiple binary classification models. The optimized XGBoost model based on ECFP4 and PubChem fingerprints achieved an ROC-AUC of 0.996 and an average precision (AP) of 0.991 on the independent test set. After that, 20 candidate compounds with high predictive probability score were finally selected through subsequent screening of the DrugBank database. Among them, DB12168 (MK-0557), DB15040 (TP-271), and DB18152 (Zilurgisertib) exhibited favorable binding free energies and stable complex formation with IRAK4 through molecular dynamics simulations and MM-PBSA calculations. Overall, these results demonstrate that approaches incorporating machine learning and structure-based computational analysis can be useful for discovering and prioritizing potential IRAK4 inhibitor candidates.