Output Dataset: Processed Adversarial Sensitivity Maps and Trained Models Derived from TCGA-BRCA for Triple Negative Breast Cancer Analysis

This dataset contains processed outputs from a multimodal adversarial sensitivity analysis of Triple Negative Breast Cancer (TNBC) using FGSM and PGD adversarial attacks as biological probes. It includes five trained histopathology image classifiers (ResNet50 and EfficientNet variants), two gene expression classifiers (Random Forest and XGBoost), per-patient adversarial sensitivity maps for TCGA-BRCA-A2 and TCGA-BRCA-E2 cohorts, ComBat batch-corrected gene expression matrices, patient splits, statistical analysis results, pathway enrichment outputs, survival data, and all figures. Raw image and expression data are not included and must be obtained from the original public sources listed in the README. All processed outputs are derived from publicly available TCGA, GEO, BreakHis, and METABRIC datasets. No human subjects research was conducted. This study used only de-identified, publicly available cancer genomics and histopathology data from TCGA, GEO, BreakHis, and METABRIC. No new patient data was collected. No IRB approval was required.