Notebooks

This collection of demonstration notebooks provides a reproducible benchmarkfor multi-omics classification and subgraph detection using the BioNeuralNet framework.

Each workflow walks through:

• Feature selection and data preprocessing.

• Network construction (similarity graphs and phenotype-aware networks).

• Hyperparameter optimization for Graph Neural Networks (GNNs).

• Downstream tasks, including disease prediction with DPMON.

• Subgraph detection and biomarker modules in selected cohorts.

The TCGA notebooks showcase the full end-to-end pipeline on multiple cancers (BRCA, LGG, KIPAN). The TCGA-LGG notebook includes a biomarker discovery section through phenotype-associated subgraphs and driver modules.

The following notebooks are included in this guide:

• TCGA-BRCA - PAM50 Cancer Subtype Classification
• TCGA-LGG (Lower Grade Glioma) - Binary Survival Prediction
• TCGA-KIPAN - Binary Cancer Staging
• ROSMAP (Alzheimer’s Disease) Biomarker Discovery