Genetic association studies have the potential to identify causative genetic variants with small effects in complex diseases, but it is not at all clear which study designs best balance power with sample size, especially when taking into account the difficulty of obtaining a sample of the necessary structure. The 14 contributions from the Genetic Analysis Workshop 15 group 3 used data sets with rheumatoid arthritis as primary phenotype from problem 2 (real data) and Problem 3 (simulated data) to investigate design and analysis problems that arise in candidate-gene, candidate-region, and genome-wide association studies. We identified three major themes that were addressed by multiple groups: (1) comparing family-based and case-control study designs with each other and with hybrid designs incorporating both related and unrelated individuals; (2) exploring and comparing techniques of combining information from multiple, correlated single-nucleotide polymorphisms; and (3) comparing analyses that select the model(s) of best fit with the ultimate aim of detecting the joint effects of several unlinked single-nucleotide polymorphisms. These contributions achieved some success in improving upon existing methods. For example, tests using related cases and unrelated controls can achieve higher power than the tests using unrelated cases and unrelated controls. Aside from these successes, the group 3 contributions highlight some interesting areas for future research.
(c) 2007 Wiley-Liss, Inc.