gms | German Medical Science

For the analysis of affected sib pairs (ASPs), a variety of test statistics is applied in genome wide scans using microsatellite markers. Even in multipoint analyses, these statistics might not fully exploit the power of a given sample, as they do not account for incomplete informativity of an ASP. For meta-analyses of linkage and association studies, it has been shown recently that weighting by informativity increases statistical power. Adopting this idea, the first aim of this paper is therefore to introduce a new class of tests for ASPs based on the mean test. To take into account how much informativity an ASP contributes, we weight families inversely proportional to their marker informativity. The weighting scheme is obtained by using the de Finetti representation of the distribution of identity by descent values. We derive the limiting distribution of the weighted mean test and demonstrate the validity of the proposed test. We show that it is extremely more powerful than the classical mean test in the case of low marker informativity. In the second part of the paper, we propose a Monte-Carlo simulation approach for evaluating significance in ASPs. We demonstrate the validity of the simulation approach for both the classical and the weighted mean test. Finally, we illustrate the use of the weighted mean test by re-analyzing two published data sets. In both applications, the maximal LOD score of the weighted mean test is 0.6 higher than the classical mean test.

Grant sponsor: DFG, grant number: ZI 591/12-1