Machine learning-guided drug repurposing for EGFR inhibition using scaffold-split validation, docking, and molecular...

Area of Science:

• Computational chemistry and pharmacology
• Drug discovery and repurposing
• Oncology and cancer research

Background:

• Aberrant epidermal growth factor receptor (EGFR) signaling drives cancer progression.
• Existing EGFR inhibitors face limitations including resistance, relapse, and toxicity.
• Novel therapeutic strategies are needed to overcome these challenges.

Purpose of the Study:

• To apply an integrated computational workflow for drug repurposing against mutant EGFR.
• To prioritize approved compounds from DrugBank for potential efficacy against resistant EGFR mutations.
• To identify novel therapeutic candidates through machine learning, docking, and molecular dynamics.

Main Methods:

• Curated EGFR bioactivity data and employed machine learning (ExtraTrees) for potency prediction.
• Utilized SwissADME descriptors and Morgan fingerprints for compound representation.
• Performed structure-based docking and molecular dynamics simulations against mutant EGFR (PDB: 6LUD).

Main Results:

• ExtraTrees model demonstrated robust performance in predicting EGFR inhibitor potency (R²=0.71).
• Docking identified Abemaciclib, Crizotinib, and Avapritinib as top drug repurposing candidates.
• Molecular dynamics simulations refined rankings, highlighting Crizotinib and Avapritinib for stability.

Conclusions:

• Machine learning-guided drug repurposing is a viable strategy for identifying novel cancer therapies.
• Computational methods can effectively prioritize compounds for experimental validation against mutant EGFR.
• This approach offers a practical pathway to overcome limitations of current EGFR inhibitors.