Survival prediction of lung cancer patients by integration of clinical and molecular features using machine learning

Abstract

Among all kinds of cancer, lung cancer has the greatest fatality rate. The mutation in Epidermal growth factor receptor (EGFR) is a significant cause of cancer deaths. Lung cancer is often diagnosed at advanced cancer stages. In this work, we propose a model by integrating patient's personal information and molecular features using machine learning classifiers, and molecular dynamics simulation. The clinical information is taken from various published studies, and molecular features are extracted using the drug-protein interactions and binding free energy of the drug-protein complex. The proposed model achieves good accuracy with a random forest classifier and a deep neural network. We believe that the prediction can be a promising index, and may help physicians and oncologists to develop personalized therapies for lung cancer patients.