Alexandre A. Todorov1, E. Genin2, K.D. Siegmund 3, P.A.F. Madden1, A.C. Heath1
Evidence is accumulating for an important genetic contribution to smoking and other drug involvement, prompting a search for study designs that would be appropriate for the detection of gene(s) underlying vulnerability to drug dependence. Work in the area of linkage analysis has clearly illustrated the difficulty to detect genes of moderate effects using traditional sampling methods. Affected sibpair (ASP) methods lose tremendous power in the presence of gene-environment interactions (A.A.Todorov, K.D. Siegmund, E. Genin, D.C. Rao, A.C. Heath, 1997, Genet. Epid. 14,541). For QTLs, sampling extremely discordant sibpairs does make many genes amenable to analysis, but often proves difficult to implement. In the present, we address the issue of sampling design in the presence of gene-environment correlations. To illustrate the principles outlined in this talk, we first consider the problem of sampling ASPs when the effect of the genes is modulated by the degree of exposure, of which the investigator has an imperfect assessment. As a second illustration, we consider the problem of determining sample sizes for a linkage study of alcohol metabolism, in which the researcher actively manipulates the degree of exposure to the drug (e.g., here, with a challenge dose of alcohol) and has the possibility to oversample some subpopulations defined by important covariates.
Address: Department of Psychiatry, Washington University School of Medicine, 40N Kingshighway, Suite 1, St.Louis MO 63139, USA, todorov@matlock.wustl.edu, (phone) 314-286-2301, (fax) 314-286-2213
1Dept. of Psychiatry, Washington University School of Medicine, St.Louis MO 63110 2Dept. of Integrative Biology, USC School of Medicine, Los Angeles CA 90033 3Dept. of Preventive Medicine, UC Berkeley, Berkeley CA 94720 Supported in part by grants AA-00728, DA-00272.