Genome-wide association studies are the hottest thing in human genetics right now, but how well do they work? The idea is to look at genetic variants in hundreds or thousands of patients with a particular disease, as well as a healthy control population. Genetic variants that show up in the patient population more often than in the control population are said to be associated with the disease, although how these variants contribute to the disease requires much more detailed follow-up.

Genome-wide association studies aren't just for studying diseases; they can be used to look for genes underlying any trait you can reliably and consistently measure, like height. The key assumption underlying these studies is that a common set of genetic variants is responsible for the trait; meaning, a few key genetic variants can explain a population's susceptibility to diabetes, or account for people's height. If this assumption - that a set of common variants can explain the differences in height or disease among a group of people - is wrong, then genome-wide association studies, which are resource intensive, are toast. Before these studies really got rolling, critics argued that there are probably too many rare genetic variants involved in determining whether you're likely to come down with a given disease, which would mean these studies wouldn't have enough statistical power to find anything.

For example, recent genome-wide studies look for genetic variants that affect human height found some variants - but those variants accounted for less than 3% of the natural variation in human height. We know, based on other types of genetic studies, that inheritance plays a large role in determining how tall you are, so this means these association studies are missing many of the genetic determinants of height.

In spite of the critics, the bandwagon rolled ahead, and people are still jumping on. A slew of these genome-wide association studies has been published recently, and now scientists can step back and see how these studies have done. A recent review in Nature Reviews Genetics takes a critical look at the first generation of genome-wide association studies. Looking at what's been achieved so far, the reviewers are reasonably optimistic about the potential of genome-wide association studies - there do seem to be some common genetic variants that are linked with common diseases, the kinds of variants that are most readily picked up by these studies.

We've found variants linked with diabetes, Crohn's disease, inflammatory bowel disease, cancer, asthma, heart disease, atrial firbrillation, cholesterol and triglyceride levels, and many more things. One group found a genetic variant that had an impact on susceptibility to both diabetes and prostate cancer. We're still a long way from having this mean anything in when you're actually in your doctors' office, but in the mean time this work is laying the foundation for a burgeoning genetic testing industry.

Links between a disease and rare genetic variants (variants that are, for example, found in only 1% of the population), are still nearly impossible to detect in these studies, which is unfortunate, because those types of rare variants probably play a disproportionately large role in disease. Individual studies don't have enough statistical power to detect them. The most important way to get around this problem, argue the reviewers, is for researchers to share their data: the data from genome-wide association studies should be available in an easily accessible form, so that researchers can combine data from multiple studies, generating enough statistical power to find rare genetic variants linked with certain traits, like height or predisposition to a disease.

So, genome-wide association studies may have worked well for some traits, not so well for others, but since these things cost so much, we should get as much mileage out of them as possible, by doing what publicly funded scientists should do anyway: share data.