2.3.3. QSAR Model Building and Optimization

The model building was conducted using Python 3.8.8 with a computer (MacOS Monterey 12.0.1, CPU intel Core-i5 2.3 GHz). Models were imported from scikit-learn and XGBoost package.^40,41 LOOCV was used to avoid overfitting and tune the hyperparameters because of our small data set size.^8,14,33 The hyperparameters with the best performance from LOOCV were used as the final model for performance evaluation with the test data set.