iCEMIGE: Integration of CEll-morphometrics, MIcrobiome, and GEne biomarker signatures for risk stratification in breast cancers

Figure 1 A schematic illustration for the study design. Using an advanced unsupervised representation learning neural network, iCEMIGE realizes efficient and effective multi-modal biomarker mining and extraction, ensuring the optimal integration of reconstructable individual biomarkers.

Figure 2 Prognostic value of the cellular morphometric biomarker signature. A: Multivariate Cox regression analysis with the hazard ratio (HR) represented as a forest plot for cellular morphometric biomarkers; B: Kaplan-Meier curves on overall survival for breast cancer patients are presented with respect to the cellular morphometric prognosis score (CMPS) groups; C: Multivariate Cox regression analysis with hazard ratio (HR) represented as a forest for CMPS groups, clinical factors, and PAM50 subtypes; D: Multivariate Cox regression analysis with the HR represented as a forest plot for CMPS, MAPS, and GEPS.

Figure 3 iCEMIGE significantly outperforms cellular morphometric prognosis score, 15-microbe abundance prognosis score, and cellular morphometric prognosis score in prognosis prediction in the Cancer Genome Atlas breast cancer cohort. A: Kaplan-Meier overall survival (OS) curves for breast cancer (BC) patients are presented according to iCEMIGE score groups; B: ROC curves for 10-year OS prediction across different signature scores. C: Area under the curve (AUC) of 10-year OS prediction across different signature scores; D: Kaplan-Meier progress-free survival (PFS) curves for BC patients are presented according to iCEMIGE score groups; E: Receiver operating characteristic (ROC) curves for 10-year PFS prediction across different signature scores. F: AUC of 10-year PFS prediction across different signature scores. The Kaplan-Meier p-values were calculated by the log-rank test among the three groups. The P values for AUC were obtained from Kruskal-Wallis test.

Figure 4 Prognostic value of iCEMIGE score on overall survival and progress-free survival according to ER status and tumor stage. A: Kaplan-Meier curves on overall survival (OS) (top panel) and progress-free survival (PFS) (bottom panel) for ER+ and ER- breast cancer (BC) patients are presented according to iCEMIGE score groups; B: Kaplan-Meier curves on OS (top panel) and PFS (bottom panel) for Stage I, II, and III&IV BC patients are presented according to iCEMIGE score groups. The P values were obtained from the log-rank test among the three groups.

Figure 5 Prognostic value of iCEMIGE scores on overall survival and progress-free survival within different molecular subtypes. Kaplan-Meier curves on overall survival (top panel) and progress-free survival (bottom panel) for breast cancer patients are presented with respect to the iCEMIGE score groups in different molecular subtypes. The P values were calculated by the log-rank test among the three groups.

Figure 6 iCEMIGE score provides significant and additional value for overall survival prediction. A: Multivariate Cox regression analysis of overall survival (OS) with hazard ratio represented as a forest for iCEMIGE score, clinical factors, and PAM50 subtypes; B: Nomogram for predicting OS was constructed based on integrating clinical factors and molecular subtype with iCEMIGE; C: C-index comparison for OS in different nomogram models with and without iCEMIGE. The P value was calculated by Mann-Whitney non-parametric test.