Analisis Komputasi Inhibitor DGAT dengan Model QSAR Menggunakan Deskriptor RDKit dan XGBoost

Abstract

Diacylglycerol acyltransferase (DGAT) is a key therapeutic target for metabolic diseases. This study aims to develop a robust Quantitative Structure-Activity Relationship (QSAR) model to predict the activity of DGAT inhibitors using a computational approach. A dataset of 197 DGAT inhibitors with pIC50 values was curated from the ChEMBL database. Molecular descriptors were calculated using RDKit, followed by feature selection through the elimination of zero-variance descriptors and correlation analysis (threshold r > 0.80), resulting in 41 representative descriptors for modeling. A predictive model was built using the Extreme Gradient Boosting (XGBoost) algorithm. Model evaluation demonstrated excellent performance with a coefficient of determination (R²) of 0.885 on training data and 0.758 on test data, alongside low RMSE and MAE values. Feature importance analysis revealed that NOCount and MaxPartialCharge descriptors were the most significant contributors to inhibitory activity, providing valuable insights for rational drug design. Although the model showed high predictive accuracy, a negative Q² from cross-validation indicates a need for further optimization. In conclusion, this study successfully demonstrates that the integration of RDKit descriptors and the XGBoost algorithm can yield an accurate and interpretable QSAR model to accelerate the insilico discovery of DGAT inhibitors.