S-GMAS: Genome-Wide Mediation Analysis With Brain Subcortical Shape Mediators

Area of Science:

• Neuroscience
• Genetics
• Medical Imaging

Background:

• Mediation analysis is crucial for understanding neurological pathways from genetic factors to clinical outcomes using brain imaging.
• Challenges exist in genome-wide mediation analysis involving numerous genetic variants (SNPs) and complex brain shape data.

Purpose of the Study:

• To develop a novel genome-wide mediation analysis framework tailored for brain subcortical shape mediators.
• To address challenges of high-dimensional genetic data, nonlinear shape spaces, and statistical inference in mediation analysis.

Main Methods:

• A fast genome-wide association analysis identifies significant genetic variants.
• Square-root velocity functions represent brain subcortical shapes in a linear Hilbert subspace.
• A shape mediation analysis framework (shape-on-scalar and scalar-on-shape models) with bootstrap resampling is employed.

Main Results:

• The proposed framework effectively handles large-scale genetic data and complex shape mediators.
• It enables robust statistical inference for direct and indirect effects in mediation pathways.
• The analysis identified significant global and spatial mediation effects.

Conclusions:

• The developed framework provides a powerful tool for genome-wide mediation analysis in neuroscience.
• It facilitates the investigation of genetic influences on clinical outcomes via brain structure.
• Application to Alzheimer's Disease Neuroimaging Initiative data demonstrates its utility.