MosGraphFlow: a novel integrative graph AI model mining signaling targets from multi-omic data

Area of Science:

• Computational Biology
• Bioinformatics
• Artificial Intelligence

Background:

• Multi-omic datasets offer a comprehensive view of cellular signaling but integrating them for biomarker discovery and pathway inference is challenging.
• Identifying key disease biomarkers and understanding complex signaling networks are crucial for developing effective therapeutic strategies.

Purpose of the Study:

• To develop a novel graph artificial intelligence (AI) model, mosGraphFlow, for analyzing multi-omic signaling graphs (mosGraphs).
• To apply the model to Alzheimer's Disease (AD) multi-omic datasets for biomarker identification and pathway analysis.
• To create a visualization tool for understanding disease-associated signaling biomarkers and networks.

Main Methods:

• Development of a novel graph AI model named mosGraphFlow.
• Analysis of multi-omic mosGraph datasets specifically for Alzheimer's Disease.
• Implementation of a visualization tool to interpret identified biomarkers and signaling networks.

Main Results:

• The mosGraphFlow model achieved superior classification accuracy compared to existing methods.
• The model successfully identified key Alzheimer's Disease biomarkers and significant signaling interactions.
• The visualization tool effectively highlighted signaling sources at specific omic levels, aiding in understanding disease pathogenesis.

Conclusions:

• The developed mosGraphFlow model provides an effective approach for integrative multi-omic data analysis.
• The model facilitates the identification of disease biomarkers and the elucidation of signaling pathways, with potential applications beyond Alzheimer's Disease.
• The publicly accessible code and visualization tool support further research in multi-omic data-driven studies.