Adaptive Sparse Multiple Canonical Correlation Analysis With Application to Imaging (Epi)Genomics Study of...

Area of Science:

• Computational Biology
• Genomics
• Neuroimaging

Background:

• Integrating multiple datasets like imaging and (epi)genomics is challenging.
• Sparse Multiple Canonical Correlation Analysis (SMCCA) is used for cross-modality correlation but suffers from unfair covariance combination.
• This unfairness limits SMCCA's feature selection power.

Purpose of the Study:

• To propose a novel formulation of SMCCA, termed adaptive SMCCA, to address the unfair combination of pairwise covariances.
• To enhance feature selection capabilities in multi-modal data integration.
• To improve the detection of three-way or multi-way correlations.

Main Methods:

• Developed adaptive SMCCA by introducing adaptive weights for combining pairwise covariances.
• Evaluated the method using simulations and real-world data analysis.
• Performed large-scale numerical experiments to assess convergence and performance.

Main Results:

• Adaptive SMCCA demonstrates superior feature selection performance compared to conventional SMCCA and fixed-weight SMCCA.
• Numerical experiments confirm adaptive SMCCA converges as rapidly as conventional SMCCA.
• Application to schizophrenia imaging (epi)genetics identified significant variants and brain regions, including novel brain-development pathways.

Conclusions:

• Adaptive SMCCA effectively overcomes the limitations of conventional SMCCA by fairly combining pairwise covariances.
• The method shows significant potential for multi-omics and imaging data integration.
• Adaptive SMCCA can uncover imaging-epigenetic associations potentially missed by traditional methods, particularly in complex biological systems.