A Novel Early-Stage Lung Adenocarcinoma Prognostic Model Based on Feature Selection With Orthogonal Regression.

ABSTRACT: Carcinoma diagnosis and prognosis are still hindered by the lack of effective prediction model and integration methodology. We proposed a novel feature selection with orthogonal regression (FSOR) method to resolve predictor selection and performance optimization. Functional enrichment and clinical outcome analyses with multi-omics information validated the method's robustness in the early-stage prognosis of lung adenocarcinoma. Furthermore, compared with the classic least absolute shrinkage and selection operator (LASSO) regression method [the averaged 1- to 4-years predictive area under the receiver operating characteristic curve (AUC) measure, 0.6998], the proposed one outperforms more accurately by 0.7208 with fewer predictors, particularly its averaged 1- to 3-years AUC reaches 0.723,

SUBMITTER: Tang B 

PROVIDER: S-EPMC7874010 | biostudies-literature | 2020

REPOSITORIES: biostudies-literature