GCNCC

GCNCC applies Graph Convolutional Networks (GCNs) within a geometric deep learning framework to integrate gene expression and protein interaction network data for identification of robust prognostic and diagnostic disease markers.

Key Features:

• Integration of Omics Data: GCNCC integrates gene expression profiles with protein interaction networks to generate network-based representations.
• Deep Network Representations: It employs GCNs to learn latent node similarity measures within the protein interaction network.
• Clustering and Classification: GCNCC clusters network nodes based on learned similarities and applies supervised feature selection to isolate clusters with high predictive power for disease-state classification.
• Cross-platform Robustness: The approach has been benchmarked across independent datasets from psychiatric disorders and cancer and across microarray and RNA-seq platforms.
• Superior Performance: GCNCC demonstrates improved accuracy and reproducibility relative to other state-of-the-art methods.

Scientific Applications:

• Genomics and Proteomics: Identification of network-based biomarkers from integrated gene expression and protein interaction data.
• Diagnostic marker discovery: Isolation of diagnostic markers that discriminate disease states across datasets and platforms.
• Prognostic marker discovery: Identification of prognostic markers associated with disease progression and outcomes.
• Cross-disease comparative studies: Comparative marker identification across disease contexts including psychiatric disorders and cancer.

Methodology:

GCNCC uses a geometric deep learning approach employing Graph Convolutional Networks to integrate gene expression with protein interaction networks, learns latent node similarity measures, performs node clustering and supervised feature selection for classification and marker identification, and is benchmarked across independent microarray and RNA-seq datasets including psychiatric disorders and cancer.