Machine Learning Predicts the Oxidative Stress Subtypes Provide an Innovative Insight into Colorectal Cancer

Figure 1

Workflow of the study. Three datasets (COAD, GSE39583, and GSE 28621) were obtained from TCGA and GEO including 1399 OS-related genes selected from the GeneCards database. Before DEGs analyses, the batch effect was removed. And nonnegative matrix factorization (NMF) was used to perform an unsupervised cluster. Hub genes were selected by Lasso regression and Cox regression to construct the signature model. DEGs, robust rank aggregation, protein-protein interaction networks were used to select hub genes to predict OS subtypes by random forest algorithms.