Trimming, weighting, and grouping SNPs in human case-control association studies.

The search for genes underlying complex traits has been difficult and often disappointing. The main reason for these difficulties is that several genes, each with rather small effect, might be interacting to produce the trait. Therefore, we must search the whole genome for a good chance to find these genes. Doing this with tens of thousands of SNP markers, however, greatly increases the overall probability of false-positive results, and current methods limiting such error probabilities to acceptable levels tend to reduce the power of detecting weak genes. Investigating large numbers of SNPs inevitably introduces errors (e.g., in genotyping), which will distort analysis results. Here we propose a simple strategy that circumvents many of these problems. We develop a set-association method to blend relevant sources of information such as allelic association and Hardy-Weinberg disequilibrium. Information is combined over multiple markers and genes in the genome, quality control is improved by trimming, and an appropriate testing strategy limits the overall false-positive rate. In contrast to other available methods, our method to detect association to sets of SNP markers in different genes in a real data application has shown remarkable success.