Using Deep Neural Network to Predict Drug Sensitivity of Cancer Cell Lines

Abstract

High-throughput screening technology has provided a large amount of drug sensitivity data for hundreds of compounds on cancer cell lines. In this study, we have developed a deep learning architecture based on these data to improve the performance of drug sensitivity prediction. We used a five-layer deep neural network, named as DeepPredictor, that integrated both genomic features of cell lines and chemical information of compounds to predict the half maximal inhibitory concentration on the Cancer Cell Line Encyclopedia (CCLE) dataset. We demonstrated the performance of our deep model using 10-fold cross-validations and leave-one-out strategies and showed that our model outperformed existing approaches.