P1.062. Discovery of Novel Dual-Target Inhibitors for EGFR and PIK3CA From Natural Products via Machine Learning and Molecular Simulation

Abstract

Abstract Topic Esophageal Cancer: Molecular Biology/Pathology Background With the advancement of personalized medicine, multi-target drug development has garnered significant attention, particularly for complex diseases such as cancer. This study aims to identify potential dual-target inhibitors against Epidermal Growth Factor Receptor (EGFR) and Phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA), two proteins whose aberrant activation is closely associated with tumorigenesis and progression in various cancers. Methods We collected IC50 values of active compounds for EGFR and PIK3CA from the BindingDB database, which were then standardized to pIC50 values using RDKit. A total of 2048 Extended-Connectivity Fingerprints (ECFPs) were calculated to serve as molecular descriptors. Various machine learning models, including Support Vector Machine (SVM), Decision Tree, Random Forest, Gradient Boosting, K-Nearest Neighbors, and LightGBM, were developed. The optimal model parameters were determined using ten-fold cross-validation and grid search, and model performance was assessed by Mean Absolute Error (MAE), Mean Squared Error (MSE), and the R-squared (R2) value. Results The SVM model demonstrated the best performance and was selected to predict activities for both EGFR and PIK3CA. Conclusion The natural product compounds CNP0456830 and CNP0467494 exhibited the lowest binding free energies for both EGFR and PIK3CA, identifying them as the most promising dual-target inhibitors. This study offers a new direction and a potential therapeutic strategy for personalized drug design in cancer treatment.