Michael B. Miller¹

The value of selected pairs of dizygotic (DZ) twins for linkage analysis is widely recognized. Monozygotic (MZ) twins are not useful for linkage analysis because they are genetically identical at every locus. However, the special value of MZ twins for studies of genotype-phenotype association in genetically complex traits has not been adequately appreciated or exploited. For quantitative traits, the strength of genotype-phenotype association (and the statistical power of the association method) is determined by the proportion of variance in the trait that is contributed by the single locus under investigation (i.e., the heritability of the locus = h₁²). We can effectively increase this heritability by using mean scores of pairs of MZ twins as the phenotype instead of using scores of singletons. Assuming no epistasis at the locus under consideration, when the overall heritability of a score is h², the locus heritability of the mean score for a pair of MZ twins is given by h₁² 2/(1+h²+c²) which is always greater than h₁² and less than 2h₁². Thus the power of association studies is improved by using mean scores for MZ pairs instead of scores of singletons, but genotyping costs hold constant. Association analysis of binary (affection status) traits can be made more powerful by using concordant MZ pairs instead of singletons. This is true regardless of method (case control, haplotype relative risk or transmission disequilibrium test [TDT]). Risch and Merikangas (1996, Science, 273, 1516-1517) computed sample sizes (singleton families and affected sib families) needed for the TDT to produce 80% power for detection of a susceptibility locus under a two-allele multiplicative model of epistasis. Under their model, a single parameter () determined the size of the effect for a locus. I demonstrate that when concordant affected MZ twins are used instead of affected singletons, the effect size is increased from to ², and parental heterozygosity is increased for low frequency penetrance-increasing alleles. As a result, the required number of families is less than for singletons or for affected sib pairs. The cost is reduced further with MZ twins because it is only necessary to genotype one twin, just as with singletons, but unlike affected sibs where both must be genotyped. Also, fewer MZ twins than sib pairs would have to be screened phenotypically to achieve any sample-size goal because of the stronger phenotypic correlation of MZ twins.

Address:   Michael B. Miller, Department of Psychology, 210 McAlester Hall, University of Missouri, Columbia, MO 65211, e-mail: mbmiller@taxa.psyc.missouri.edu, web: http://taxa.psyc. missouri.edu/~mbmiller/

¹Department of Psychology, University of Missouri, Columbia, MO

BGA 1998 Home Page