A study of feasibility for genome-wide haplotype association of complex traits in imaging genetics

Abstract

Imaging-genetics cohorts allow to find associations between genotyped variants and brain imaging features. They are invaluable tools to evaluate the part of genetics and environment in the brain characteristic variance observed in normal and pathological populations. The present analyses were conducted using the 2018 UK Biobank (UKB) data release, and included 15,040 subjects for which sulcal opening, a measure of sulcus width, are extracted for 126 sulci using BrainVisa/Morphologist pipeline. Based on genetic maps, continuous blocks of high-confidence phase are extracted using the haplotype dataset of UK Biobank. The feasibility study analyzes 13,942 haplotypes of 1,756 blocks on chromosome 21, obtained with a stringent block definition, and shows that block-based test seems underpowered compared to haplotype-based association test.

Genome-wide haplotype association study in imaging genetics using whole-brain sulcal openings of 16,304 UK Biobank subjects

“…As a compromise between haplotype length and uncertainty, we chose δ value equal to 0.001 cM, leading to 119,548 haplotype blocks across the whole genome (see Supplementary Material 4 and [21]). For individual SNP phase, UK Biobank estimated a median switch error rate of 0.229% on chromosome 20 taken as an example [6], corresponding to a median value of 37 or 38 individual SNPs with phase errors.…”

“…We studied a range of values for the δ parameter from 10 −3 to 2.5 10 −2 cM, that balance (i) the size of haplotype blocks, by including SNPs further apart on the map, while ensuring (ii) the reliability of the phase along the block [21]. We set the δ parameter to the smallest value (smallest probability of For each run of permutation, tests were corrected by the Bonferroni procedure P T = α ∕N T for N T tests with risk α = 0.05 (see section "Genomewide significance threshold").…”

“…(Right) Distributions of L(h) (block length in cM), as defined by the genetic distance between bordering markers of each block. Blocks were defined over the 22 autosomal chromosomes using δ = 0.001 cM, a parameter that account for the maximal length in cM between two consecutive markers in a same haplotype (see section "Data processing" and[21] for details). The distribution of block lengths is homogeneous across the 22 autosomal chromosomes.…”