SmulTCan: A Shiny application for multivariable survival analysis of TCGA data with gene sets

Abstract

Background Survival analysis is widely used in cancer research, and although several methods exist in R, there is the need for a more interactive, flexible, yet comprehensive online tool to analyze gene sets using Cox proportional hazards (CPH) models. The web-based Shiny application (app) SmulTCan extends existing tools to multivariable CPH models of gene sets—as exemplified using the netrins and their receptors (netrins-receptors). It can be used to identify survival gene signatures (GSs) and select the best subsets of input gene, microRNA, methylation level, and copy number variation sets from the Cancer Genome Atlas (TCGA). Objectives To create a tool for CPH model building and best subset selection, using survival data from TCGA with input gene expression files from UCSC Xena. Furthermore, we aim to analyze the input TSV file of netrins-receptors in SmulTCan and discuss our findings. Methods SmulTCan uses Shiny's reactivity with built-in R functions from packages for CPH model analysis and best subset selection including “survminer”, “riskRegression”, “rms”, “glmnet”, and “BeSS”. Results Results from the SmulTCan app with the netrins-receptors gene set indicated unique hazard ratio GSs in certain renal and neural cancers, while the best subsets for this gene set, obtained via the app, could differentiate between prognostic outcomes in these cancers.