Figure 4. Feature selection of risk score model. (A) Selection of tuning parameters in the Lasso regression analysis based on 1,000 cross-validations. (B) Lasso regression analysis coefficients. (C) The importance of XGBoost machine learning screening the top 30 genes. (D) Multivariate analysis of 3 genes (COCH, MYOM2, PDE1B) and establishment of the regression equation. (E–G) Kaplan-Meier curve analysis of the relationship between the expression levels of COCH, MYOM2 and PDE1B, respectively, and the prognosis of OSs patients.