Strategic Multi-Omics Data Integration via Multi-Level Feature Contrasting and Matching.

Area of Science:

• Bioinformatics
• Data Science
• Computational Biology

Background:

• Multi-omics data analysis is crucial but challenging due to data sparsity, high dimensionality, and heterogeneity.
• Integrating diverse omics sources requires sophisticated methods to extract meaningful biological insights.

Purpose of the Study:

• To develop an advanced model, MFCC-SAtt, for effective feature extraction and integration of multi-omics data.
• To address the challenges of data sparsity, high dimensionality, and heterogeneity in omics data analysis.

Main Methods:

• MFCC-SAtt employs a multi-level feature contrast clustering approach with self-attention mechanisms.
• Autoencoders with self-attention are used for each omics modality to compress features into a shared space.
• A multi-level feature extraction framework and semantic information extractor mitigate optimization conflicts.

Main Results:

• MFCC-SAtt demonstrates exceptional performance in multi-omics data analysis.
• In a pan-cancer clustering task, the model achieved an accuracy exceeding 80.38%.

Conclusions:

• MFCC-SAtt effectively integrates and extracts informative features from multi-omics data.
• The model enhances deep clustering quality and offers a robust solution for complex biological data challenges.