Prediction of P2Y12 antagonists using a novel genetic algorithm-support vector machine coupled approach

Area of Science:

• Computational Chemistry
• Medicinal Chemistry
• Pharmacology

Background:

• P2Y12 receptor antagonists are crucial in treating thrombotic disorders.
• Accurate prediction of P2Y12 inhibition activity is vital for drug development.
• Existing quantitative structure-activity relationship (QSAR) studies for P2Y12 antagonists are limited.

Purpose of the Study:

• To develop and validate a novel computational model for predicting P2Y12 antagonist activity.
• To optimize molecular descriptor subsets for enhanced QSAR model performance.
• To compare the proposed model against other established machine learning methods.

Main Methods:

• A genetic algorithm (GA)-support vector machine (SVM) coupled approach was employed.
• Kernel-based nonlinear projection was utilized to enhance SVM performance.
• The GA-SVM model was compared with GA-Partial Least Squares (PLS), GA-Random Forest (RF), and GA-Gaussian Process (GP) models.

Main Results:

• The GA-SVM model achieved high prediction accuracy with R²=0.976, Rcv²=0.829 (training), and Rpred²=0.811 (test set).
• The proposed GA-SVM model significantly outperformed GA-PLS, GA-RF, and GA-GP methods.
• The model demonstrated excellent internal and external predictive capacity for P2Y12 antagonists.

Conclusions:

• The GA-SVM approach is a powerful and efficient tool for predicting P2Y12 antagonist activity.
• This model can aid in the screening and optimization of novel P2Y12 antagonists.
• The findings support the use of this computational strategy in early-stage drug discovery.