Machine Learning-Based Virtual Screening and Identification of the Fourth-Generation EGFR Inhibitors

Area of Science:

• Oncology
• Pharmacology
• Computational Chemistry

Background:

• Advanced non-small cell lung cancer (NSCLC) treatment faces challenges with the emergence of the EGFR C797S mutation, rendering existing inhibitors ineffective.
• The need for novel therapeutic strategies is critical to overcome resistance in EGFR-mutated NSCLC.

Purpose of the Study:

• To identify innovative fourth-generation EGFR tyrosine kinase inhibitors (EGFR-TKIs) using machine learning techniques.
• To develop a predictive model for uncovering new EGFR-TKI candidates against resistant mutations.

Main Methods:

• A fusion framework combining the full quadratic effect model and Lasso model was employed for critical molecular descriptor selection from high-dimensional, sparse data.
• Machine learning models were developed to predict the efficacy of potential EGFR-TKIs.
• Virtual screening was performed to identify promising hit compounds.

Main Results:

• Novel small-molecule inhibitors were designed and synthesized based on structural descriptors.
• The synthesized inhibitors demonstrated potent activity against double and triple mutated EGFR kinases (L858R/T790M/C797S and Del19/T790M/C797S).
• Machine learning-driven virtual screening successfully identified four hit compounds with potential therapeutic value.

Conclusions:

• Machine learning approaches can effectively accelerate the discovery of next-generation EGFR-TKIs.
• The identified hit compounds warrant further investigation for their therapeutic potential in treating resistant NSCLC.
• This study provides a framework for developing novel EGFR-TKIs against challenging mutations like C797S.