Framework for identifying quantitative trait loci in association studies using structural equation modeling

In this article, we suggest a framework for identifying quantitative trait loci (QTL) in association studies using structural equation modeling. Two t ests to detect QTLs and estimate the proportion of variance they explain ar e discussed. The first test assumes that there is no population admixture a nd only requires that the subjects are genotyped. The second one is a TDT-l ike test that cannot give false-positive results due to population admixtur e but requires that the parents of the subjects are genotyped as well and t hat subjects have at least one heterozygous parent. Power calculations show ed that with the first test, 100 subjects were generally sufficient to dete ct a locus that explained 10% and less than 1,000 subject to detect a locus that explained 1% of the total variance. To obtain the same power, the TDT -like test required an initial sample that was on average 1.7 times larger. Calculations showed that the first test was quite robust against populatio n admixture and that the power of tests to detect admixture was good. This suggested that in the extreme and very specific conditions in which populat ion admixture may cause false-positive findings, admixture can often be det ected. (C) 2000 Wiley-Liss, Inc.