Genome-wide interaction analysis reveals replicated epistatic effects on brain structure.

Area of Science:

• Neurogenetics
• Brain Imaging
• Computational Biology

Background:

• Alzheimer's disease risk is influenced by genetic factors, prompting research into common genetic variants (SNPs).
• Genome-wide analysis of SNP-SNP interactions is computationally intensive due to the vast number of potential tests.
• Machine learning, specifically iterative sure independence screening, enables analysis of datasets with more predictors than observations.

Purpose of the Study:

• To perform a genome-wide interaction analysis of all possible SNP-SNP interactions.
• To identify SNP-SNP interactions affecting regional brain volumes measured by magnetic resonance imaging (MRI) and tensor-based morphometry.
• To investigate the association of identified interactions with brain volume and potential protective effects.

Main Methods:

• Utilized an implementation of the sure independence screening algorithm (EPISIS) for genome-wide interaction analysis.
• Tested all possible SNP-SNP interactions influencing regional brain volumes.
• Validated findings in the Alzheimer's Disease Neuroimaging Initiative (ADNI) and an independent twin imaging dataset (Queensland Twin Imaging).

Main Results:

• Identified a significant SNP-SNP interaction between rs1345203 and rs1213205, explaining 1.9% of the variance in temporal lobe volume.
• Mapped voxelwise effects of the interaction across the whole brain in both ADNI and Queensland Twin Imaging datasets.
• Each unit increase in interaction loading correlated with approximately 5% greater regional brain volume, indicating a protective effect.

Conclusions:

• A specific SNP-SNP interaction significantly influences temporal lobe volume.
• This interaction demonstrates a protective effect, associated with increased regional brain volume.
• The findings highlight the potential of advanced machine learning methods for uncovering complex genetic interactions in neuroimaging studies.