GIANT: an online resource for comprehensive survival analysis in pan-cancer from The Cancer Genome Atlas (Preprint)
BACKGROUND Prognostic genes or gene signatures have been widely used to predict patients’ survival and aid the decision of therapeutic options. Although few web-based survival analysis tools to identify them have been developed, they only provide limited information. OBJECTIVE To overcome limitations of previous web-based tools and provide comprehensive survival analysis, we developed GIANT, an online resource for identifying prognostic biomarkers in pan-cancer from The Cancer Genome Atlas (TCGA). METHODS We used R program to code survival analysis based on RNA-seq data from TCGA (n=10,320). To perform survival analyses, we excluded patients and genes that have insufficient information (survival status, tumor stage, age, gender, cancer type, blast count, and histologic grade). The GIANT is programmed by applying appropriate cross validation methods and survival analysis methods to provide three analysis services (survival analysis by single gene, cancer type, variable signature). RESULTS It can perform comprehensive survival analysis to identify prognostic genes or gene signatures with reflecting tumor heterogeneity. Using RNA-seq, clinical data and pathway databases in combination, it provides gene/variable signature by grouped variable selection methods (least absolute shrinkage and selection operator, Elastic Net regularization, Network-Regularized high-dimensional Cox-regression) that has better discriminatory power than single gene. Users also can find prognostic values of gene and statistically significant genes in specific cancer. All results are presented as Kaplan-Meier curve with median/optimal cutoff value, C-index, and area under the curve (AUC) value at t-years. Moreover, users can easily obtain results in the forms of graphs and tables. CONCLUSIONS In conclusion, the GIANT has made it possible to easily perform integrated survival analysis while overcoming the limitations of previous online tools. It will help scientists of those who are vulnerable to computer technology to do database analysis can easily perform comprehensive survival analysis.