TEMINET: A Co-Informative and Trustworthy Multi-Omics Integration Network for Diagnostic Prediction

Area of Science:

• Biomedical investigation
• Computational biology
• Genomics

Background:

• Integrated multi-omics data aids in understanding complex human diseases.
• Challenges exist in perceiving intra- and inter-omics informativeness due to intricate relations.
• Existing multi-omics integration methods face difficulties.

Purpose of the Study:

• To introduce TEMINET, a novel multi-omics integration approach.
• To enhance diagnostic prediction for complex human diseases.
• To address challenges in multi-omics data integration.

Main Methods:

• TEMINET utilizes an intra-omics co-informative representation module.
• A trustworthy learning strategy is employed for inter-omics fusion.
• Graph attention networks and a multi-level framework are applied to construct disease-specific networks.
• A combined-beliefs fusion approach harmonizes omics representations.

Main Results:

• TEMINET demonstrates advanced performance in classification tasks across four diseases.
• The approach shows robustness in diagnostic prediction.
• Experiments utilized mRNA, methylation, and miRNA data.

Conclusions:

• TEMINET offers an effective solution for multi-omics integration.
• The method improves the understanding of complex disease mechanisms.
• TEMINET enhances diagnostic prediction accuracy and reliability.