Research Paper Volume 15, Issue 7 pp 2667—2688
Machine learning-based construction of immunogenic cell death-related score for improving prognosis and response to immunotherapy in melanoma
Figure 4. Construction of ICDscore. (A) Venn diagram to screen DEGs with a significant prognostic p-value < 0.1 in the TCGA-SKCM and GSE65904 cohorts. (B, C) The LASSO Cox regression model was constructed from 182 DEGs, and the tuning parameter (λ) was calculated based on the partial likelihood deviance with ten-fold cross validation. 4 signature genes were identified according to the best fit profile. (D–G) Kaplan–Meier curves of OS in melanoma patients from ICDscore-high and ICDscore-low subgroups of GSE65904 (D), TCGA-SKCM (E), GSE54467 (F), and GSE22153 (G) datasets. (H) Clinical characteristics and RNA expression level of 4 crucial genes in melanoma patients from ICDscore-high and ICDscore-low subgroups of the TCGA-SKCM dataset. (I) Univariate analysis of ICDscore and other clinical characteristics in TCGA-SKCM dataset. (J) Multivariate analysis shows ICDscore, breslow depth, M stage and age were independent prognostic factors. Ns, not significant; *p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001.