Construction of IRAK4 inhibitor activity prediction model based on machine learning

Area of Science:

• Medicinal Chemistry
• Computational Biology
• Drug Discovery

Background:

• Interleukin-1 receptor-associated kinase 4 (IRAK4) is vital in Toll-like receptor and Interleukin-1 receptor signaling pathways.
• IRAK4's role in immunity, inflammation, and cancer makes it a key target for small-molecule inhibitors.
• Accurate prediction of IRAK4 inhibitor activity is crucial for efficient drug development.

Purpose of the Study:

• To develop and validate a predictive model for IRAK4 inhibitor activity.
• To screen a large chemical library for potential IRAK4 inhibitors.
• To identify novel drug-like compounds targeting IRAK4.

Main Methods:

• A machine learning model was built using LightGBM and molecular fingerprints on a dataset of 1628 IRAK4 inhibitors.
• High-throughput molecular docking screened 1.6 million compounds from the chemdiv database.
• ADMET predictions and drug-likeness rules were applied to filter potential candidates.
• Molecular dynamics simulations were used to confirm binding stability.

Main Results:

• The predictive model achieved R²=0.829, MAE=0.317, and RMSE=0.460 on independent testing.
• The model successfully predicted activities for 90 IRAK4 inhibitors from 2023.
• 34 compounds were identified as promising IRAK4 inhibitor candidates after rigorous screening and validation.
• Molecular dynamics confirmed stable binding of the identified compounds to IRAK4.

Conclusions:

• A robust machine learning model for IRAK4 inhibitor activity prediction has been established.
• This study identified 34 novel compounds with potential as IRAK4 inhibitors.
• The findings provide valuable insights for structure-guided design of new IRAK4 inhibitors.