DeepProg: an ensemble of deep-learning and machine-learning models for prognosis prediction using multi-omics data

Background Prognosis (survival) prediction of patients is important for disease management. Multi-omics data are good resources for survival prediction, however, difficult to integrate computationally.

Results We introduce DeepProg, a new computational framework that robustly predicts patient survival subtypes based on multiple types of omic data. It employs an ensemble of deep-learning and machine-learning approaches to achieve high performance. We apply DeepProg on 32 cancer datasets from TCGA and discover that most cancers have two optimal survival subtypes. Patient survival risk-stratification using DeepProg is significantly better than another multi-omics data integration method called Similarity Network Fusion (p-value=7.9e-7). DeepProg shows excellent predictive accuracy in external validation cohorts, exemplified by 2 liver cancer (C-index 0.73 and 0.80) and five breast cancer datasets (C-index 0.68-0.73). Further comprehensive pan-cancer analysis unveils the genomic signatures common among all the poorest survival subtypes, with genes enriched in extracellular matrix modeling, immune deregulation, and mitosis processes.

Conclusions DeepProg is a powerful and generic computational framework to predict patient survival risks. DeepProg is freely available for non-commercial use at: http://garmiregroup.org/DeepProg