Tuesday, July 18, 2006

Common disease, common variant   posted by JP @ 7/18/2006 01:53:00 PM

In the years after whole-genome association scanning was proposed as a more powerful alternative to linkage studies for locating the alleles involved in "complex" (i.e. involving multiple genetic and environmental factors) phenotypes, there was much debate about the true prospects of this approach. One of the major points of contention in this debate centered around what is now called the CDCV (common disease, common variant) hypothesis, which proposed that common polymorphisms would confer susceptibility to common diseases like heart disease, diabetes, etc.

Why was this an important question? If common polymorphisms were underlying these traits, it would explain the inability of linkage studies to find them, as linkage studies have low power in this situation. In this case, association studies might succeed in places where linkage studies could never come up with any consistent results (see linkage versus association). However, if complex diseases were caused by multiple rare loci, the association approach would still be powerless and, well, human geneticists would be out of luck. For population genetic models supporting either side (multiple rare variants versus common variants) see the links in this post.

Five to ten years on, what's the emerging consensus? Well, as is often the case in these kinds of scientific debates, both sides are kind of right. In some cases, rare variants play a role, and in others, common variants. But it's clear that those who argued against the whole-genome association approach and the development of the HapMap on the grounds that the CDCV hypothesis was unfounded are currently in the process of ending up looking a little Luddite.

A new paper in Nature Genetics gives another example of a common disease being influenced by a common polymorphism. In this case, the risk of a heart attack in a Japanese population was shown to be slightly increased by the presence of a mutation, present on 35% of chromosomes, in a gene that regulates inflammation. The increase in risk is slight-- on the order of a 20% increase in the probability of having a heart attack, but significant nonetheless.

This suggests where population genetic models in the past have gone wrong: common disease phenotypes are products of underlying variation in quantitative traits, and assigning a "fitness" to a certain level of a trait is extremely difficult, as different genetic backgrounds and environmental variation can absolutlely swamp small effects. So assuming a polymorphism that causes a disease must be detrimental, as opposed to neutral, nearly neutral or even beneficial, creates a model that may not track reality. As I've argued before, new paragims for describing alleles that influence disease are arising; "associated with disease = negative fitness" is out.