Construction and validation of gene signature for prognosis and drug sensitivity in cholangiocarcinoma based on cellular senescence related genes

Cholangiocarcinoma is a very deadly epithelial cell cancer with poor clinical outcome. Cellular senescence plays a vital role in the oncogenesis and the aggressiveness of cholangiocarcinoma. Integrative machine learning procedure including 10 methods (random survival forest, elastic network, Lasso, Ridge, stepwise Cox, CoxBoost, partial least squares regression for Cox, supervised principal components, generalized boosted regression modelling, and survival support vector machine) was performed to construct a cellular senescence-related signature (CSS) for cholangiocarcinoma. Cellular experiment was used to verify the biological function of hub gene. The optimal prognostic CSS developed by Lasso method served as an independent risk factor and had a stable and powerful performance in predicting the overall survival rate in cholangiocarcinoma, with the AUC of 1-, 3-, and 5-year ROC curve being 0.957, 0.929 and 0.928 in TCGA cohort. Low CSS score indicated with a lower tumor immune dysfunction and exclusion score, lower tumor microsatellite instability score, lower immune escape score, lower MATH score, and higher tumor mutation burden score in cholangiocarcinoma. Down-regulation of EZH2 inhibited the proliferation, colony and promoted apoptosis of cholangiocarcinoma cell. Integrative machine learning analysis was performed to construct a novel CSS in cholangiocarcinoma. This CSS acted as an indicator for predicting the prognosis and immunotherapy benefits of cholangiocarcinoma patients.