A generalised threshold defined case-control selection strategy for QTL association mapping in pooled DNA samples

Disequilibrium mapping is an essential tool in the identification of genes underlying complex traits. DNA pooling is a technique that allows one to perform a case-control association study without genotyping the individual DNA samples, but rather a single pool of cases and a single pool of controls. Selecting from the extremes of the trait distribution for analysis greatly enhances power in a sample. However, from where in the trait distribution should we select the two pools, to achieve optimal power in a sample? Through analytical derivation and optimisation, we provide robust guidelines for optimal pooling fractions. Our studies suggest that in the absence of experimental error a symmetric sampling scheme of genotyping individuals from the 20-30% tails of the trait distribution is optimal under a variety of possible genetic models underlying the trait of interest. We also consider the effects of experimental noise from sources such as measurement error and unequal DNA contribution from individuals. We show that a standard error of up to 1% in allele frequency estimation can be tolerated without great loss in power, but the optimal symmetric pooling fraction is decreased to 10-20%.