Identification of signature of tumor-infiltrating CD8 T lymphocytes in prognosis and immunotherapy of colon cancer by machine learning

To explore the specific marker of CD8+ T cell subsets which are closely related to the prognosis and immunotherapy of patients with colon cancer. 18 kinds of immune cell expression profile data sets were obtained from GEO database. Compared with other immune cell types, the specific markers of CD8 (+) T cells (TI-CD8) in colorectal cancer were screened. Regression analyses were used to further screen prognostic related genes and construct a prognostic evaluation model. The patients were stratified and analyzed according to the risk scores, KRAS mutation status, stage, lymphatic infiltration and other indicators. The landscape of infiltration level, mutation and copy number variation of immune subsets in high and low TI-CD8Sig score groups were compared and analyzed. The difference of drug response between high and low TI-CD8Sig score groups was analyzed. Differential expression of the model genes was verified by the HPA database. Six prognostic-related CD8T cell-specific gene targets were further screened, and the prognostic evaluation model was constructed. The AUC value of the model is >0.75. FAT3 and UNC13C showed a high mutation state in the low-risk group, while USH2A, MUC5B et al. specifically showed a high mutation state in the high-risk group. Compared with the low-risk group, the high-risk group had lower effective rate of drug response. The expression of PD-1 gene was positively correlated with the level of TI-CD8Sig score. The risk assessment model based on CD8T cell-specific marker genes can effectively predict the prognosis and the drug response of patients with CRC.