A powerful test of parent-of-origin effects for quantitative traits using haplotypes

Abstract

Imprinting is an epigenetic phenomenon where the same alleles have unequal transcriptions and thus contribute differently to a trait depending on their parent of origin. This mechanism has been found to affect a variety of human disorders. Although various methods for testing parent-of-origin effects have been proposed in linkage analysis settings, only a few are available for association analysis and they are usually restricted to small families and particular study designs. In this study, we develop a powerful maximum likelihood test to evaluate the parent-of-origin effects of SNPs on quantitative phenotypes in general family studies. Our method incorporates haplotype distribution to take advantage of inter-marker LD information in genome-wide association studies (GWAS). Our method also accommodates missing genotypes that often occur in genetic studies. Our simulation studies with various minor allele frequencies, LD structures, family sizes, and missing schemes have uniformly shown that using the new method significantly improves the power of detecting imprinted genes compared with the method using the SNP at the testing locus only. Our simulations suggest that the most efficient strategy to investigate parent-of-origin effects is to recruit one parent and as many offspring as possible under practical constraints. As a demonstration, we applied our method to a dataset from the Genetics of Lipid Lowering Drugs and Diet Network (GOLDN) to test the parent-of-origin effects of the SNPs within the PPARGC1A, MTP and FABP2 genes on diabetes-related phenotypes, and found that several SNPs in the MTP gene show parent-of-origin effects on insulin and glucose levels.

Figure 1. Power of detecting parent-of-origin effect in nuclear families using the testing locus only.

Three causal minor allele frequencies (MAFs) are considered and are shown is three separate pannels. Four family sizes are considered and the numbers of siblings 1–4 are indicated by different colors red, green, brown, and blue. Solid lines are for no missing (complete) parental genotypes, and dotted lines are for one parent's genotype missing.

Figure 2. Power of detecting parent-of-origin effect using different haplotype block lengths for different family sizes.

The length of haplotype blocks 1–5 are indicated by different colors blue, red, green, purple, and brown. They are compared with the power using the variance components (vc) method, shown by grey color.

Figure 3. Power of detecting parent-of-origin effect with and without missing genotypes.

Three scenarios are considered: no missing, (i) missing one parent's genotype in each family, (ii) for 10% random genotype missing, shown in blue, red, and green, respectively. Results connected by solid lines are obtained by the method using haplotypes, and the results connected by dotted lines are obtained by the method using the testing locus only.