Construction of a gene signature and the evaluation of its prognosis and prediction

In order to evaluate the value on the prognosis of m6A-related genes and develop a potential risk model, Lasso cox regression analysis was utilized on their expression in the TCGA dataset [24, 25]. Univariable cox analysis was performed to screen out the genes that were related to survival. The minimum criterion was set as p value less than 0.05, then four genes and their coefficients were determined. And λ as the best penalty parameter related to the TGGA dataset were selected. The equation was used to compute the risk score of the signature [26]:

In which Coefficients represents the coefficient, while xi is the representative of relative expression value of the Z-score transformation of every chosen regulator. Each patient’s risk score was calculated by this formula in TCGA dataset. In LC cases, high-risk group (the risk score of these samples exceeds 0.9539055) and low-risk group (the risk score of these samples is inferior to 0.9539055) were determined on the strength of the risk score of the tumor samples. Moreover, a nomogram was established, which assimilated the four selected genes with LC prognosis and we conducted 3-year and 5-year ROC (receiver operating characteristic curve) analysis to assess the nomogram. In addition, Cox regression analyzed the clinical characteristics correlated with the overall survival rate of LC patients with univariate and multivariate analysis, and we applied the Kaplan–Meier method to assess the practicality of risk prognostic models. Further, we conducted ROC analysis to detect the sensitivity and specificity of risk score.