Multimodal Omics Data Integration Using Max Relevance--Max Significance Criterion.

Area of Science:

• Bioinformatics
• Computational Biology
• Data Science

Background:

• High-dimensional omics data presents challenges for feature extraction.
• Existing methods may lack efficiency or accuracy in multimodal analyses.

Purpose of the Study:

• To present CuRSaR, a novel supervised regularized canonical correlation analysis method.
• To extract relevant and significant features from multimodal high-dimensional omics datasets.

Main Methods:

• CuRSaR integrates regularized canonical correlation analysis (RCCA) with a rough hypercuboid approach.
• An analytical formulation based on spectral decomposition relates CCA and RCCA.
• A hypercuboid equivalence partition matrix computes feature relevance and significance.

Main Results:

• CuRSaR demonstrates significantly lower computational complexity compared to existing methods.
• The method achieves superior classification accuracy on real-life data.
• The equivalence partition matrix efficiently optimizes regularization parameters.

Conclusions:

• CuRSaR offers an effective and efficient approach for feature extraction in multimodal omics data.
• The method advances the analysis of complex biological datasets.