The invention discloses gene markers for forecasting an immunotherapy effect on a cancer and application of the gene markers. The markers comprise BRD4, RPS6KA4, TET1, BCL10, FGFR2, RAD51C, TCF7L2, TGFBR1, AR, BBC3, H3C11, TSC2, NSD1, PNRC1, IGF2, H3C3, SOCS1, TCF3, TSHR, PMS2, EGFR and CDKN2A. According to the application, a model constructed by 22 genes can be used for predicting an immunotherapy effect and better instructing clinical medication, and the susceptibility and specificity of prediction of the model are superior to those of PD-L1 expression.