Construction of a prognostic model of lung adenocarcinoma based on machine learning

Abstract

In order to more accurately predict the prognosis and survival of lung adenocarcinoma patients, this paper used the gene expression and clinical information data of lung adenocarcinoma patients in the open database of TCGA to jointly construct a prognosis model of lung adenocarcinoma. Three difference analysis methods and univariate cox regression analysis were used as the preliminary screening method. By comparing the variable selection ability of lasso regression and random survival forest, comparing the performance of cox proportional risk regression model and random survival forest model, and integrating clinical data, a model that can more accurately predict the prognosis of lung adenocarcinoma patients was constructed. After comparison and selection, lasso regression was used to select variables and cox proportional risk model was used as the prediction model. The consistency index of the model reached 0.712. The AUC for 1-year, 3-year and 5-year survival of lung adenocarcinoma patients in the validation set were 0.808, 0.816 and 0.754, respectively. After the fusion of clinical data, the 1-year, 3-year and 5-year survival prediction AUC in the validation set were 0.840, 0.836 and 0.865, respectively, indicating that the model had good predictive performance.