mosGraphGen: a novel tool to generate multi-omics signaling graphs to facilitate integrative and interpretable graph...

Area of Science:

• Computational Biology
• Bioinformatics
• Artificial Intelligence in Medicine

Background:

• Multi-omics data (genomics, epigenomics, transcriptomics, proteomics) offer a holistic view of cellular signaling.
• Integrating and interpreting multi-omics data for biomarker discovery remains a significant challenge.
• Graph AI models are suitable for multi-omics analysis but require pre-processed, biologically meaningful graph data.

Purpose of the Study:

• To develop a method for generating biologically meaningful multi-omics signaling graphs from raw multi-omics data.
• To simplify the process of applying graph AI models to multi-omics datasets.
• To facilitate the identification of critical biomarkers from integrated multi-omics data.

Main Methods:

• Developed mosGraphGen (multi-omics signaling graph generator) to create sample-specific multi-omics signaling graphs (mos-graphs).
• Mapped multi-omics data onto a pre-existing multi-level signaling network.
• Normalized data by aggregating measurements and aligning to a reference genome.

Main Results:

• Successfully generated mos-graphs from multi-omics data.
• Demonstrated the utility of mosGraphGen using The Cancer Genome Atlas (TCGA) and Alzheimer's disease (AD) datasets.
• Enabled direct application and evaluation of graph AI models by researchers.

Conclusions:

• mosGraphGen effectively converts complex multi-omics data into a format suitable for graph AI analysis.
• The tool addresses the challenge of pre-analyzing and structuring multi-omics data for AI models.
• Facilitates advanced biomarker discovery and understanding of cellular signaling pathways.