gms | German Medical Science

The case-only (CO) design is a powerful approach to study gene-environment (GxE) interactions using only affected subjects. Genotype imputation uses a reference sample such as the Haplotype Reference Consortium (HRC) to predict genotypes at untyped loci. However, using healthy controls as a reference in a CO study may introduce systematic error, especially in regions of genetic main effects. We investigated the effects of imputation accuracy and choice of imputation base on the validity and power to detect GxE interactions in a CO design.

Using data from 719 Crohn's Disease (CD) cases from Kiel, Germany, we investigated the imputation accuracy for target SNPs with varying minor allele frequencies (MAFs) with and without genetic main effects. Target SNPs were imputed using neighboring proxy SNPs with different levels of linkage disequilibrium (LD) and the HRC as a reference base. True genotypes of target SNPs were available for comparison. Furthermore, we simulated different levels of GxE interaction by assigning environmental exposure conditional on the SNP genotype in order to evaluate the loss in statistical power.

The comparison of true and imputed MAFs of target SNPs showed that imputation accuracy depends on local LD structure, but also on presence of a genetic main effect. The highest differences between true and imputed MAFs were of SNPs found on chromosome 16 in the areas of IL27 and NOD2 genes, which are known to play a role in CD. Here, the estimated MAF based on imputed genotypes was more similar to the MAF in controls than to the true MAF in the case-only sample. This leads to a varying loss of statistical power to detect GxE, which can be dramatic for the imputed target SNPs with main effects.

In conclusion, our study describes constellations in which imputed data should be used with caution when testing for GxE interactions in CO studies.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.