Where is the causal variant? On the advantage of the family design over the case-control design in genetic association studies.

Many associated single-nucleotide polymorphisms (SNPs) have been identified by association studies for numerous diseases. However, the association between a SNP and a disease can result from a causal variant in linkage disequilibrium (LD) with the considered SNP. Assuming that the true causal variant is among the genotyped SNPs, other authors demonstrated that the power to discriminate between it and other SNPs in LD is low. Here, we propose to take advantage of the information provided by family data to improve the inference on the causal variant: we exploit the linkage information provided by affected sib pairs to discriminate the causal variant from the associated SNPs. The family-based approach improves discrimination power requiring up to five times less individuals than its case-control equivalent. However, the main advantage of family design is the possibility to carry out the procedure one step further: the linkage information allows inference on causal variants, which are not genotyped but in LD with tag-SNPs displaying association, which is impossible with case-control design. By means of Bayesian methods, we estimate the LD between the observed SNPs and an unobserved causal variant, as well as the allelic odds ratio at the unobserved causal variant. The proposed procedure is illustrated on a multiple sclerosis (MS) family data set including genotypes of SNPs in IL2RA, confirming the advantage of using a family design to identify causal variants. The results of our method on this data suggest the existence of two distinct causal variants in this gene for the MS.