With interval mapping (i.e., multipoint linkage analysis) of an illness gene (2 flanking marker loci), the power of linkage detection is increased as the density of markers increases up to approximately a lOcM map, but not beyond. If heterogeneity of linkage is present, interval mapping is only efficient if the analysis allows for heterogeneity. Appropriate tests and criteria have been determined for excluding linkage of a disease locus to a map. This is critical when screening the whole genome. We found that there is a large increase in power to detect a linkage to a map when calculating the lod score allowing for heterogeneity (lod2) compared with the standard lod score assuming homogeneity (lod1) even when increasing the lod score criterion to allow for the heterogeneity parameter. Lod2 is substantially more powerful than lod1 for not rejecting a true linkage. We have computed the expected linkage information in our sample of bipolar families by simulation for different levels of marker information and for different recombination fractions. For example, the average lod score per family for a completely informative marker linked at 1% recombination is > 2.0. For a marker with 3 alleles, the average lod is 1.3. The increase in the type I error of linkage analysis when carrying out multiple hypothesis testing on the same genotype data has been investigated. For example, if lod scores are calculated under 2 different genetic models and 3 different sets of classification criteria for illness, the type I error is increased by 3-4 fold, indicating that the lod score criterion needs to be increased.
