RF-Lung-DR: Integrating Biological and Drug SMILES Features in a Random Forest-Based Drug Response Predictor for Lung Cancer Cell Lines

Thi-Oanh Tran, Quang-Hien Kha, Nguyen-Quoc-Khanh Le

Abstract: In the era of precision medicine, predicting drug responses accurately is crucial for tailoring patient-specific treatments. Despite advances in machine learning (ML) models for drug response prediction (DRP), challenges remain in predicting effective therapies with high accuracy. This study introduces RF-Lung-DR, a ML model that integrates biological markers and drug SMILES features to predict drug responses in lung cancer cell lines, thus enhancing drug screening processes. Using drug sensitivity data from the Genomics of Drug Sensitivity in Cancer (GDSC), the model was developed across seven ML algorithms, with Random Forest (RF) proving to be the most effective for optimizing DRP accuracy. RF-Lung-DR achieved prediction accuracies of 80% in lung squamous cell carcinoma (LUSC) and 78% in lung adenocarcinoma (LUAD). The investigation also identified key biological biomarkers and drug SMILES features that significantly influence predictive performance. Focusing on lung cancer-a leading cause of cancer-related mortality worldwide-RF-Lung-DR’s methodology supports the broader application of personalized medicine and underscores the potential for developing individualized patient care strategies in oncology.