Extended Data Fig. 6: Survival and recurrence analyses on TCGA and QHCG dataset, and the correlation maps of clinical parameters.
From: Deep learning supported discovery of biomarkers for clinical prognosis of liver cancer
a, b, Kaplan-Meier analyses of patient stratification of low and high death risk patients via M2MNet on TCGA dataset (a) and QHCG dataset (b). c-f, Kaplan-Meier analyses of patient stratification of low and high recurrence risk patients via M2MNet (c), MacroNet (d), TND (e), and NEC (f) on QHCG dataset. g-i, Multivariable analyses of factors associated with recurrence and MacroNet (g), TND (h), and NEC (i) on QHCG dataset (n = 83 patients); the data are presented as hazard ratio estimates (squares) and the error bars show the 95%-confidence interval of the hazard ratio estimate, according to multivariable Cox proportional hazards model. The results of univariate, multivariate analyses, and the abbreviations of each variable are detailed in Supplementary Table 3. j, k, Correlation maps of clinical parameters on TCGA dataset (j) and QHCG dataset (k). P values according to two-sided log-rank test (a-f) and multivariable Cox proportional hazards model (g-i). n, sample size; HR, hazard ratio; Stage, AJCC staging; TIL, tumor infiltrating lymphocytes digital score; BDT, bile duct thrombosis; AFP, alpha-fetoprotein; MVI, microvascular invasion.
