Research Paper Volume 15, Issue 20 pp 11244—11267
CDKN2A was a cuproptosis-related gene in regulating chemotherapy resistance by the MAGE-A family in breast cancer: based on artificial intelligence (AI)-constructed pan-cancer risk model
Figure 9. Artificial intelligence identified CRG subgroups. (A) Processing of identifying 44 target genes, and they are put into AI training. In this part, 75% of pan-cancer data are defined as a training cohort, while the last 25% is defined as a testing cohort. Six types of AI functions are performed, including XGboost, RandomForest, Deep-Learning, SVM, Multi-logistics, and KNN. (B) Prognosis differences amongst CRG subgroups which are identified by AI. Then, the GEO breast cancer cohort (GSE58812, GSE42568, GSE20685, n=538) is divided into four CRG subgroups by XGboost, and here displayed 44-gene expression profile (C) and its disturbance differences (D) (the representatives selected in the black box have statistical differences). Single-gene-mediated prognosis signature is explored (E). CIBERSORT explored immune cell infiltration features in CRG subgroups (F), in which the representatives selected in the black box have statistical differences. (G) K-M analysis showed prognosis differences amongst four CRG subgroups in breast cancer.