Recently, the Type-1 Diabetes Genetics Consortium (T1DGC) conducted a genome-wide linkage scan of 2,658 affected sib-pairs with type-1 diabetes [18]. Using formulae outlined by Risch and Merikangas [19] it is straightforward to assess the power of this study to detect linkage given a model specifying the number of risk variants and their effect sizes. Figure 3 compares, under various models consistent with synthetic association, the power of the T1DGC linkage scan to the power of a GWAS with 2,000 cases and 2,000 controls [6]. When individual odds ratios are ≥3, or there are many independent risk variants, linkage (rather than association) is the more powerful approach. If synthetic associations are common, this observation yields a testable prediction: either GWAS signals should overlap substantially with results from well-powered linkage scans, or all synthetic associations arise from the relatively small parameter space where linkage is poorly powered but GWAS is not.
