TY - JOUR T1 - netDx: Interpretable patient classification using integrated patient similarity networks JF - bioRxiv DO - 10.1101/084418 SP - 084418 AU - Shraddha Pai AU - Shirley Hui AU - Ruth Isserlin AU - Muhammad A Shah AU - Hussam Kaka AU - Gary D. Bader Y1 - 2017/01/01 UR - http://biorxiv.org/content/early/2017/09/26/084418.abstract N2 - Patient classification has widespread biomedical and clinical applications, including diagnosis, prognosis and treatment response prediction. A clinically useful prediction algorithm should be accurate, generalizable, be able to integrate diverse data types, and handle sparse data. Importantly, the resulting model should be easily interpretable, as clinicians are unlikely to trust black box statistical models. We describe netDx, the first supervised patient classification framework based on patient similarity networks. netDx meets the above criteria and particularly excels at data integration and model interpretability. We demonstrate the features of this framework by integrating up to six heterogeneous datatypes, including clinical variables, DNA methylation, somatic mutations, mRNA, miRNA and protein expression profiles, for survival prediction in kidney, lung, ovarian and brain cancer. As a machine learning tool, netDx outperforms eight standard machine-learning methods in predicting binary survival in renal clear cell carcinoma, and performs at par with these methods in predicting ovarian carcinoma. In comparison to traditional machine learning-based patient classifiers, netDx results are more interpretable, visualizing the decision boundary in the context of patient similarity space and identifying biological pathways and other features important for prediction. Using pathway-level features in predicting kidney cancer survival from transcriptome data, netDx identified known and potentially novel kidney cancer pathways and biomarkers. Thus, netDx can serve both as a useful classifier and as a tool for discovery of biological features characteristic of disease. Upon publication, an open-source R/Java implementation of netDx will be made publicly available along with sample files and automation workflows packaged as vignettes. ER -