Aside from a few major successes, there have been many problems replicating significant associations between polymorphic gene variation and complex diseases. There are several reasons for this, of which population structure is widely considered to be the most important. Population structure will affect both the validity and power of experiments and may be particularly important when relative risks are slight or the alleles involved are rare. With low relative risks and/or rare alleles, sample sizes need to be much larger than those often used in case-control studies and as sample size increases, the amount of population structure needed to perturb the results decreases. To address this problem, there are several statistical methods available that attempt to allow for the effect of population structure to be taken into account.

However, these methods are not really satisfactory and so the only suitable alternative is to design the studies with greater care and a powerful approach may be to characterise genetically both the cases and controls. Individuals from the controls can then be chosen to match the cases so as to minimise the stochastic differences between the two populations. We are therefore assembling a UK control population as a resource for future studies. It will comprise samples from 3500 individuals, who will have been carefully selected from throughout the UK. Rural regions will be targeted to avoid the admixture observed in large urban environments and volunteers will be sought who were born in the same place as their parents and grandparents to ensure historical integrity. The collection will be genotyped for around 3000 markers, with the aim of identifying about 200 ancestrally informative markers (AIMs). These AIMs will then be used to match controls to cases.