GN-SCCA: GraphNet based Sparse Canonical Correlation Analysis for Brain Imaging Genetics.

Area of Science:

• Neuroimaging Genetics
• Statistical Genetics
• Computational Neuroscience

Background:

• Identifying associations between genetic variants and neuroimaging quantitative traits (QTs) is crucial in brain imaging genetics.
• Sparse Canonical Correlation Analysis (SCCA) is a common technique for uncovering complex multi-SNP-multi-QT associations.
• Existing SCCA methods sometimes struggle to effectively incorporate prior biological knowledge and covariance structures.

Purpose of the Study:

• To propose a novel structured SCCA method, GraphNet, for enhanced discovery of genetic variant-neuroimaging trait associations.
• To integrate prior knowledge through a Graph constrained Elastic-Net regularizer for coefficient smoothness.
• To incorporate covariance structure information often overlooked by conventional SCCA.

Main Methods:

• Development of a novel structured SCCA method utilizing a Graph constrained Elastic-Net regularizer.
• Incorporation of prior knowledge to induce smoothness between adjacent coefficients in a graph structure.
• Integration of covariance structure information into the SCCA model.

Main Results:

• The proposed GraphNet method demonstrated superior performance compared to a widely used SCCA method on both simulated and real imaging genetic data.
• GraphNet successfully identified important genetic associations with neuroimaging QTs.
• The method yielded biologically interpretable findings, facilitating a deeper understanding of genetic influences on brain structure and function.

Conclusions:

• The novel GraphNet method offers an effective approach for brain imaging genetics research.
• Integrating graph-based prior knowledge and covariance structures enhances the discovery of genetic associations.
• GraphNet provides interpretable results, advancing the field of neuroimaging genetics.