MOSDNET: A multi-omics classification framework using simplified multi-view deep discriminant representation learning...

Area of Science:

• Computational biology and bioinformatics
• Systems biology
• Genomics, transcriptomics, proteomics, and metabolomics

Background:

• Integrating multi-omics data is crucial for understanding complex diseases.
• Current methods struggle to effectively extract shared and specific representations from diverse omics datasets.
• Accurate disease classification and biomarker identification remain significant challenges.

Purpose of the Study:

• To introduce MOSDNET, a multi-omics classification framework designed to extract both shared and specific data representations.
• To enhance disease classification accuracy and efficiency using advanced machine learning techniques.
• To identify pivotal biomarkers for deeper insights into disease etiology and progression.

Main Methods:

• Leveraged Simplified Multi-view Deep Discriminant Representation Learning (S-MDDR) for representation extraction with similarity and orthogonal constraints.
• Integrated multi-omics data by concatenating extracted representations.
• Employed Dynamic Edge GCN (DEGCN) with patient similarity networks to learn intricate data structures and node representations.
• Utilized a multitask learning approach for training, optimizing both data integration and classification.

Main Results:

• MOSDNET demonstrated superior classification accuracy compared to state-of-the-art multi-omics classification models in extensive comparative experiments.
• The framework successfully identified pivotal biomarkers within the multi-omics data.
• Achieved enhanced accuracy and efficiency in disease classification.

Conclusions:

• MOSDNET provides a robust and effective framework for multi-omics data integration and disease classification.
• The ability to extract shared and specific representations significantly improves classification performance.
• MOSDNET offers valuable insights into disease mechanisms through biomarker identification, aiding in the development of novel diagnostics and therapeutics.