Fig. 6

Construction of classifier and prediction of immunotherapeutic response. A Volcano plots of differentially expressed genes (DEGs) between subtype 1 and subtype 2 in the SCCH cohorts. Genes with false discovery rate (FDR) < 0.05 and |log2 fold change|> 1 are considered as DEGs. The horizontal dashed line indicates the FDR of 0.05. The vertical dashed lines indicate the log2 fold change values of -1 and 1. Blue points: down-regulated genes; Gray points: non-differential genes; Red points: up-regulated genes. B Error rate curve based on tenfold cross-validation using an increased number of variables. The vertical dashed lines indicate that the lowest error rate was obtained when the number of variables is 83, which is the best model. C ROC curves in the SCCH cohort (red line), TCGA cohort (blue line), and GEO cohort (green line). The area under the curves (AUC) was used to evaluate the accuracy of the classifier. D Concordance of cervical SCC subtypes prediction between the 83-gene-based classifier and original classification based on NMF. E Subtype 1 shares a high similarity with the PD-1 response group in all three cohorts. Subtype 2 shares a high similarity with the PD-1 no-response group in the SCCH cohort and TCGA cohort. The colors labeled in each cell indicate the Bonferroni adjusted p-values of each subclass association