A Pleiotropic Map of Brain Imaging Genetics Reveals Biologically Distinct Latent Endophenotypes

Area of Science:

• Neuroimaging genetics
• Quantitative genetics
• Computational neuroscience

Background:

• Genomewide association studies (GWAS) of brain imaging phenotypes are challenged by high dimensionality and collinearity of image-derived phenotypes (IDPs).
• Identifying the genetic architecture of complex brain traits requires methods that can handle numerous, correlated phenotypes.

Purpose of the Study:

• To introduce DIMPLE-GWAS, a novel dimensionality-reduction framework for analyzing high-dimensional, pleiotropic brain imaging phenotypes.
• To identify latent genetic structures within a large set of IDPs and improve the power of genetic discovery.

Main Methods:

• Applied a dimensionality-reduction framework (DIMPLE-GWAS) to approximately 4000 IDPs from the UK Biobank cohort (∼33K participants).
• Validated the identified latent phenotypes in an independent cohort (ABCD).
• Compared heritability and locus discovery power of latent phenotypes against individual IDPs.

Main Results:

• Identified 25 biologically interpretable latent phenotypes from ∼4000 IDPs, which were validated in an independent cohort.
• DIMPLE-GWAS clusters showed substantially greater heritability than individual IDPs.
• Discovered 104 novel genomewide-significant loci, increasing power for genetic discovery compared to prior GWAS of single IDPs.

Conclusions:

• DIMPLE-GWAS effectively reduces dimensionality in brain imaging genetics, revealing latent phenotypes with higher heritability and discovery power.
• The genetically defined latent phenotypes partially deviate from traditional brain atlases, suggesting novel patterns of brain organization.
• These findings highlight new genetic relationships between brain structure, neurologic, psychiatric, cognitive, and behavioral phenotypes.