Non-Linear Deep Neural Network for Rapid and Accurate Prediction of Phenotypic Responses to Kinase Inhibitors

Abstract

Protein kinase inhibitors are one of the most successful targeted therapies to date, with more than 50 FDA-approved drugs and hundreds in clinical development. Despite this progress additional kinase inhibitors are needed to expand the target space as well as overcome drug resistance that has emerged in clinical setting. Application of deep learning algorithms has the potential to reduce the overall costs of pre-clinical drug development and accelerate discovery of high-value lead compounds. Here, we developed KiDNN (Kinase Inhibitor prediction using Deep Neural Networks) to predict phenotypic effects of kinase inhibitors in cancer cell lines. Unlike previous studies that used linear networks to model kinase signaling, we now introduce a non-linear, multilayer feed-forward network that more closely mimic complex and dynamic nature of kinase-driven signaling pathways. We used KiDNN to predict the effect of ~200 kinase inhibitors on migration of cultured breast and liver cancer cells as a model phenotype. We compared prediction accuracy of KiDNN to other prediction tools based on linear models and determined through experimental testing that KiDNN outperformed the linear models. Further, we show that DNN hyperparameters learned from one set of data (breast cancer cells) can be used to generate KiDNN that predicts responses in multiple, unrelated cancer models with minimal training data. We validated that an inhibitor of tyrosine kinase receptors, and an inhibitor of Src family kinases decreased migration of triple-negative breast cancer (Hs578t) cells, consistent with the role of these kinases in driving motility. Overall, we show that non-linear, DNN-based models provide a powerful approach to in silico screen hundreds of kinase inhibitors. Our study further supports the potential of deep learning algorithms to reduce the overall costs and accelerate the discovery of lead compounds for subsequent development of drug candidates.