Fig. 2

Identification and Validation of Diagnostic Genes for BC using Machine Learning Algorithms. A, B Utilizing the LASSO logistic regression algorithm with penalty parameter tuning via tenfold cross-validation, 49 BC-related features were selected from the DE-MRGs (A, B). C KEGG analysis of the 49 diagnostic genes. D, E The SVM-RFE algorithm identified an optimal combination of 14 feature genes to distinguish BC samples from normal samples. F KEGG analysis of the 14 genes. G Intersection of marker genes obtained from the LASSO and SVM-RFE models revealed four common genes (GLRX2, NMT1, PPP2R2B, and TRAF3IP3) for subsequent analysis. H The interacting network of GLRX2, NMT1, PPP2R2B, and TRAF3IP3 was constructed using the Genemania database, illustrating potential functional connections. I A logistic regression model incorporating the seven marker genes demonstrated robust performance, as indicated by ROC curves. J The diagnostic value of the model was validated in the GSE3167 dataset. K The expression patterns of GLRX2, NMT1, PPP2R2B, and TRAF3IP3 in BC samples and normal samples, based on GSE13507 datasets. ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05. NS, not significant