Abstract
The arrival of highly dense genetic maps at low cost has geared the focus of linkage analysis studies toward developing methods for placing putative trait loci in narrow regions with high confidence. This shift has led to a new analytic scheme that expands the traditional twoโstage protocol of preliminary genome scan followed by fine mapping through inserting a new stage in between the two. The goal of this new โintermediateโ fine mapping stage is to isolate disease loci to narrow intervals with high confidence so that association studies can be more focused, efficient, and costโeffective. In this paper, we compared and contrasted five methods that can be used for performing this intermediate step. These methods are: the lod support approach, the generalized estimating equations (GEE) method, the confidence set inference (CSI) procedure, and two bootstrap methods. We compared these methods in terms of the coverage probability and precision of localization of the resulting intervals. Results from a simulation study considering several twoโlocus models demonstrated that the two bootstrap methods yield intervals with approximately correct coverage. On the other hand, the 1โlod support intervals, and those produced by the GEE method, tend to significantly undercover the trait locus, while the regions obtained by the CSI incline to overcover the gene position. When the observed coverage of the confidence intervals produced by all the methods was held to be the same, those obtained through the CSI procedure displayed a higher ability to localize loci, especially when these loci have a minor contribution to the trait and when the amount of data available for the analysis is relatively small. However, with very large sample sizes, lod support intervals emerged as a winner. Application of the methods to the data from the Arthritis Research Campaign National Repository led to intervals containing the position of a known trait locus for all methods, with the greatest precision achieved by the CSI. Genet. Epidemiol. 2006. ยฉ 2006 WileyโLiss, Inc.