Abstract—Breast cancer prognosis benefits from multi-omics integration but faces challenges of class imbalance, calibration, and limited interpretability. To develop a federated multi-omics framework that enhances predictive accuracy, stability, and interpretability while preserving data privacy, the framework uses a Tab Transformer (TT) backbone with Graph Capsule (GC) for structural feature learning, Golden Eagle Optimization (GEO) for stable convergence, and explainability modules Shapley Additive Explanations (SHAP), Gradient-weighted Class Activation Mapping (Grad-CAM)) for transparent outputs. Federated learning was applied across the Cancer Genome Atlas-Breast Invasive Carcinoma (TCGA-BRCA), the Cancer Genome Atlas Pan-Cancer (TCGA-PANCAN) BRCA subset, and Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) datasets. Evaluation used accuracy, precision, recall, F1-score, Matthews Correlation Coefficient (MCC), Cohen’s kappa, Top-2 accuracy, and Jensen-Shannon divergence. The global model achieved 98.7% accuracy, an F1-score of 0.983, an MCC of 0.951, and the lowest Jensen-Shannon (JS) divergence of 0.039. GC improved feature interactions, GEO enhanced optimization stability, and the explainability modules supported biologically relevant insights. The framework improved accuracy, fairness, and calibration while ensuring interpretability and privacy, showing strong potential for clinical adoption.
 
Keywords—breast cancer prognosis, multi-omics, federated learning, tab transformer, graph capsule, golden eagle optimization, explainable artificial intelligence


Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).