Fig. 5: Multivariable logistic regression analysis of DEFDRGs for keloid classification.

A Cross-validation plot depicting binomial deviance as a function of log(λ) during Lasso regression model selection. B Heatmap illustrating expression profiles of DEFDRGs in keloid and healthy samples. C Differential expression of DEFDRGs in high- vs. low-risk groups across total, train, and test sets. D Clinical correlation analysis showing associations between risk scores and demographic/clinical parameters (age, sex, diagnosis, and time). E The ROC curve of the train and test sets demonstrating the predictive performance of the DEFDRG-based model for keloid classification (train set AUC = 0.995, test set AUC = 0.868). F GSEA enrichment plots for representative signaling pathways enriched in high- and low-risk groups. DEFDRGs differentially expressed fibroblast-differentiation-related genes, ROC receiver operator characteristic, AUC area under the curve, GSEA gene set enrichment analysis.