Transferability and fine-mapping of genome-wide associated loci for adult height across human populations

Abstract

Human height is the prototypical polygenic quantitative trait. Recently, several genetic variants influencing adult height were identified, primarily in individuals of East Asian (Chinese Han or Korean) or European ancestry. Here, we examined 152 genetic variants representing 107 independent loci previously associated with adult height for transferability in a well-powered sample of 1,016 unrelated African Americans. When we tested just the reported variants originally identified as associated with adult height in individuals of East Asian or European ancestry, only 8.3% of these loci transferred (p-values < or =0.05 under an additive genetic model with directionally consistent effects) to our African American sample. However, when we comprehensively evaluated all HapMap variants in linkage disequilibrium (r(2) > or = 0.3) with the reported variants, the transferability rate increased to 54.1%. The transferability rate was 70.8% for associations originally reported as genome-wide significant and 38.0% for associations originally reported as suggestive. An additional 23 loci were significantly associated but failed to transfer because of directionally inconsistent effects. Six loci were associated with adult height in all three groups. Using differences in linkage disequilibrium patterns between HapMap CEU or CHB reference data and our African American sample, we fine-mapped these six loci, improving both the localization and the annotation of these transferable associations.

Figure 1. Principal coordinate analysis.

Shown are the first three dimensions from classical multidimensional scaling of the allele sharing distance matrix. Red represents individuals from the HapMap CEU sample, green represents individuals from the HapMap YRI sample, and blue represents individuals from the HUFS sample (African Americans). The lower panels show two-dimensional projections of the first three dimensions for the HUFS sample including the CEU and YRI reference samples. The upper panels show two-dimensional projections of the first three dimensions for just the HUFS sample. The diagonal panels show the eigenvalues and in parentheses the variance explained by the first three dimensions.

Figure 2. Genomic locations of SNPs previously associated with adult height.

Black dots indicate associations originally discovered in populations of European ancestry. Red dots indicate associations originally discovered in populations of Chinese Han ancestry. Blue dots indicate associations originally discovered in populations of Korean ancestry. The y-axis represents discovery p-values on the −log₁₀ scale. The light gray line indicates a p-value of 5×10⁻⁷.

Figure 3. Schematic diagram of different linkage disequilibrium patterns in discovery and follow-up samples.

A) The associated tag SNP and a causal variant are in the same region of linkage disequilibrium in both the discovery and follow-up samples. Both the exact and local approaches may yield successful transfer. B) The associated tag SNP and a causal variant are not in the same region of linkage disequilibrium in both the discovery and follow-up samples. The exact approach will fail for the original tag SNP but the local approach may succeed for other proxy SNPs if they are in the same region of linkage disequilibrium as the causal variant in the follow-up sample.

Figure 4. Association p-values and linkage disequilibrium in the HUFS sample for the height locus at chromosome 1p12.

The open red diamond indicates SNP rs12735613 (for which the association was discovered in individuals of European ancestry), the filled red diamond indicates rs17038182 (for which the association was discovered in individuals of Korean ancestry), and the blue diamond indicates the SNP associated in the HUFS (African American) sample. The boundaries of the set of SNPs reflect r ²≥3 in the HapMap CHB data surrounding rs17038182.

Figure 5. Association p-values and linkage disequilibrium in the HUFS sample for the height locus at chromosome 17q23.2.

The open red diamond indicates SNP rs757608 (for which the association was discovered in individuals of European ancestry), the filled red diamond indicates rs2079795 (for which the association was discovered in individuals of Korean ancestry), and blue diamonds indicate SNPs associated in the HUFS (African American) sample. The boundaries of the set of SNPs reflect r ²≥3 in the HapMap CHB data surrounding rs2079795.