Investigate the relevance of major signaling pathways in cancer survival using a biologically meaningful deep learning model

Abstract

Survival analysis and prediction are important in cancer studies. In addition to the Cox proportional hazards model, recently deep learning models have been proposed to integrate the multi-omics data for survival prediction. Cancer signaling pathways are important and interpretable concepts that define the signaling cascades regulating cancer development and drug resistance. Thus, it is interesting and important to investigate the relevance to patients survival of individual signaling pathways. In this exploratory study, we propose to investigate the relevance and difference of a small set of core cancer signaling pathways in the survival analysis of cancer patients. Specifically, we built a biologically meaningful and simplified deep neural network, DeepSigSurvNet, for survival prediction. In the model, the gene expression and copy number data of 1648 genes from 46 major signaling pathways are used. We applied the model on 4 types of cancer and investigated the relevance and difference of the 46 signaling pathways among the 4 types of cancer. Interestingly, the interpretable analysis identified the distinct patterns of these signaling pathways, which are helpful to understand the relevance of the signaling pathways in terms of their association with cancer survival time. These highly relevant signaling pathways can be novel targets, combined with other essential signaling pathways inhibitors, for drug and drug combination prediction to improve cancer patients survival time.