Discovery of New HER2 Inhibitors via Computational Docking, Pharmacophore Modeling, and Machine Learning

Area of Science:

• Computational chemistry and drug discovery
• Oncology and cancer therapeutics
• Pharmacology and medicinal chemistry

Background:

• The human epidermal growth factor receptor 2 (HER2) is a key oncogene in aggressive cancers like breast cancer.
• Developing novel HER2 inhibitors is essential for improving treatment outcomes in HER2-driven malignancies.

Purpose of the Study:

• To create a computational workflow for identifying novel HER2 inhibitors.
• To generate and validate pharmacophores for HER2 inhibition using machine learning.
• To screen a chemical database for new potential anti-HER2 drug candidates.

Main Methods:

• Flexible docking was used to generate pharmacophores from 15 known HER2 inhibitors.
• Machine learning-quantitative structure-activity relationship (ML-QSAR) models were built using 1,272 HER2 inhibitors.
• Genetic Function Algorithm (GFA) was employed for feature selection, combined with Bagging and J48Graft classifiers.

Main Results:

• The best ML-QSAR models identified two key pharmacophores (Hypo_1 and Hypo_2) crucial for HER2 inhibitor activity.
• Screening the National Cancer Institute (NCI) database yielded three promising novel anti-HER2 leads.
• The top lead compound showed an experimental anti-HER2 IC50 of 3.85 μM and possessed a unique chemical scaffold.

Conclusions:

• The developed computational workflow effectively identifies novel HER2 inhibitors.
• The identified leads represent promising candidates for further development against HER2-related cancers.
• These novel inhibitors possess distinct chemical structures, potentially overcoming resistance mechanisms associated with existing therapies.