Genetic signatures of human brain structure: A comparison between GWAS and relatedness-based regression

Abstract

Identifying the genetic variations impacting human brain structure and their further effects on cognitive functions, is important for our understanding of the fundamental bases of cognition. In this study, we take two different approaches to this issue: classical genome-wide association analysis (GWAS) and a relatedness-based regression approach (REL) to search for associations between genotype and brain structural measures of gray matter and white matter. Instead of searching genetic variants by testing the association between a phenotype trait and the genotype of each single-nucleotide polymorphism (SNP) as in GWAS, REL takes advantage of multiple SNPs within a genomic window as a single measure, which potentially find associations wherever the functional SNP is in linkage disequilibrium (LD) with SNPs that have been sampled. We also conducted a simulation analysis to systemically compare GWAS and REL with respect to different levels of LD. Both methods succeed in identifying genetic variations associated with regional and global brain structural measures and tend to give complementary results due to the different aspects of genetic properties used. Simulation results suggest that GWAS outperforms REL when the signal is relatively weak. However, the collective effects due to local LD boost the performance of REL with increasing signal strength, resulting in better performance than GWAS. Our study suggests that the optimal approach may vary across the genome and that pre-testing for LD could allow GWAS to be preferred where LD is high and REL to be used where LD is low, or the local pattern of LD is complex.