MATH: A Deep Learning Approach in QSAR for Estrogen Receptor Alpha Inhibitors

Area of Science:

• Computational chemistry
• Oncology
• Pharmacology

Background:

• Breast cancer is a leading cause of death in women, with hormone therapies targeting estrogen as a key treatment.
• Conventional drug discovery for estrogen receptor alpha inhibitors is time-consuming and expensive.

Purpose of the Study:

• To develop and validate a novel computational framework for predicting estrogen receptor alpha inhibitors.
• To leverage quantitative structure-activity relationship (QSAR) analysis combined with machine learning for drug discovery.

Main Methods:

• Utilized supervised learning integrating self-attention Transformer and molecular graph information.
• Developed five classification models to predict estrogen receptor alpha inhibitors.
• Evaluated model performance using precision, recall, F1 score, specificity, and ROC AUC.

Main Results:

• The proposed MATH model achieved high performance metrics: precision (0.952), recall (0.972), F1 score (0.960), specificity (0.922), and ROC AUC (0.977).
• MATH demonstrated robust predictive capabilities for estrogen receptor alpha inhibitors.

Conclusions:

• The MATH model shows significant potential to aid researchers in identifying candidate compounds for estrogen alpha inhibitors.
• This approach can guide and accelerate drug discovery pathways for breast cancer therapies.