Construction and verification of prognostic models
TCGA-COAD was used in the trial cohort, and GES40976 was used in the verification cohort. In the mRNA expression data in TCGA-COAD, we used univariate Cox regression analysis to obtain a total of 22 DEG related to patient survival (P Value <0.05). In the LASSO COX regression analysis, P Value <0.05 was filtered as a statistically different prognostic gene. After re-sampling 1000 times using the R language ”glmnet” software package[7], a predictive model that affects the prognosis was established. The median risk score value was selected as the cutoff value of the COAD cohort, and divided into high-risk group and low-risk group. The ”survival” software package and ”survminer” software package of R language were used to perform Kaplan-Meier analysis and draw survival curves Figure to verify the correlation between the prognostic model and overall survival. According to the different risk scores of the patients, the risk curve diagram, survival status diagram and heat map are drawn.