Structure-Based Design and Machine Learning-Driven Prioritization of EGFR Inhibitors.

Area of Science:

• Medicinal Chemistry
• Computational Drug Discovery
• Cancer Therapeutics

Background:

• Epidermal Growth Factor Receptor (EGFR) is crucial for tumor growth and survival.
• EGFR Tyrosine Kinase Inhibitors (TKIs) are effective but face resistance.
• There is a continuous need for novel and improved EGFR inhibitors.

Purpose of the Study:

• To design and identify novel EGFR inhibitors using a scaffold-based approach integrated with machine learning.
• To evaluate the binding affinity and stability of potential inhibitors against EGFR.
• To accelerate the discovery of effective and stable drug candidates for cancer treatment.

Main Methods:

• Synthesis of an 81-compound chemical library based on a urea-aryl hydrazone core.
• Utilizing machine learning (Random Forest) with molecular fingerprints (PaDEL-Descriptor) to predict inhibitory activity (pIC₅₀).
• Employing molecular docking and 100-nanosecond molecular dynamics simulations to assess binding and stability.

Main Results:

• Identified 81 unique chemical structures with potential EGFR inhibitory activity (pIC₅₀ > 6).
• Docking studies revealed stable interactions, including hydrogen bonds and hydrophobic contacts, comparable to Erlotinib.
• Molecular dynamics simulations confirmed the sustained stability of lead candidates' interactions with the EGFR kinase.

Conclusions:

• Computational methods successfully identified lead molecules with predicted activity and stable binding poses.
• Selective halogenation enhanced interaction stability, maintaining key hydrogen bonds with the EGFR hinge region.
• This integrated approach accelerates the identification and refinement of novel EGFR inhibitors for cancer therapy.