Lung cancer remains a significant global health concern, posing a substantial burden on both patients and healthcare systems. As a result, there is an urgent need for innovative therapeutic interventions to manage lung cancer more effectively. In this study, we developed classification models using machine learning algorithms to predict drug responses in lung cancer cell lines. A diverse dataset was retrieved, consisting of 692 active and 1,071 inactive compounds tested against five major lung cancer cell lines: CaLu-06, HCC-78, NCI-H322, NCI-H358 and NCI-H522. Drug-like properties of these compounds were generated and employed as descriptors for model development. The proposed method utilised techniques such as z-score, correlation analysis, recursive feature elimination with cross-validation and SMOTE to pre-process the data and identify key features. Further, hyperparameter optimisation was conducted using Optuna to fine-tune model parameters and enhance performance. The results revealed that Random Forest reached an accuracy of 0.80 and an AUC of 0.85. This positions it as the best model, with significant implications for drug discovery and personalised lung cancer therapies. The implementation materials alongside python code are accessible freely at https://github.com/Gokulakrish13/Machine-Learning-Classifiers-for-Predicting-Active Molecules-Against-Lung-Cancer-Cells.git.
Kanagasabapathy et al. (Tue,) studied this question.