Genome scans and candidate gene approaches in the study of common diseases and variable drug responses.

There is both tremendous enthusiasm and controversy surrounding genetic association studies. In this article we use new methods for representing variation in the human genome that suggest as a few as 170 000 carefully selected single nucleotide polymorphisms (SNPs) are sufficient to represent all the common SNPs in European or East Asian population samples, and up to 300 000 in West African population samples. We also show that efficient genome scans require a detailed description of genomic variation in humans, as envisioned in the HapMap project. Although these analyses are encouraging about the prospects for genome scans, we argue that the difficulty of fine-localization of causal variants will be an important obstacle to progress, and that for reasons of both cost and statistical power candidate gene approaches are likely to remain the primary focus of genetic association studies for years to come.