S. S. Cherny1,2, D. W. Fulker1,2, & P. Sham2
Powerful methods currently exist for detecting association in samples of individuals when parental genotypic data are available, allowing simultaneous control for such things as admixture and population stratification, which could otherwise yield false-positive results. However, there are no available methods for detectings allelic associations in unselected, population-based samples of sibships which allow for such statistical control. It is now widely accepted that variance components methods for mapping QTLs using sibships are optimally powerful. We propose an extension of the maximum-likelihood variance-components method of detecting linkage in siblings which incorporates a model on the phenotypic means in addition to the usual model of covariance structure conditional on allelic sharing. The method involves partitioning the mean effect of a locus into a between- and within-family component, thereby controlling for stratification and admixture, while simultaneously modelling linkage, resulting in a further increase in power in cases of weak association due to the trait locus not being in complete disequilibrium with the marker locus. Power is explored for the method under various conditions and the method is compared with a simple analysis-of-variance approach to association data.
Address: Institute for Behavioral Genetics, Campus Box 447, University of Colorado, Boulder, CO 80309-0447, Phone: +1 303 492 0835, FAX: +1 303 492 8063, Email: Stacey.Cherny@Colorado.EDU, WWW: http://ibgwww.colorado .edu/~cherny/
1Institute for Behavior Genetics, University of Colorado, Boulder, CO 80309-0447, 2 Social, Genetic and Development Psychiatry Research Centre, Institute of Psychiatry, DeCrespigny Park, Camberwell, London SE5 8AF, United Kingdom, 2Supported in part by DA-11015 and a grant from the Medical Research Council of Great Britain