Tuesday, February 06, 2007
Ever since 1996, it has been a foregone conculusion that the linkage study would be de-throned in favor of the association study as the preferred method for searching for genes involved in common human diseases. New technologies and the availablity of resources like the HapMap have now nearly made that foregone conclusion a reality. Large-scale association studies are no panacea, of course--they certainly have their limitations, and perhaps some members of the human genetics community have been too bold in their promises of what they will bring. But a new paper lauches a puzzling attack on the entire field, all in the service of demolishing a straw man.
The paper is entitled "Are genome-wide association studies all that we need to dissect the genetic component of complex human diseases?" The authors apparently think the answer is "no", and I'm certainly not going to disagree with them there-- this is the straw man. It would be nice if they could give an example of one person who thinks that genome-wide association studies are all that is needed to understand complex disease-- a dumbass grad student, a dotty old professor, anyone. But the best they can do is to quote a HapMap press release as saying that the HapMap -- and the association studies it allows -- are a "new tool that will accelerate the discovery of genes related to common disease". This statement is, well, true. Even the authors think so: several pages later, perhaps having forgotten what they wrote in the introduction, they opine that "the availability of [new] tools and information are a valued and welcome addition to the armamentarium geneticists can use to identify and characterize the genetic component to complex diseases". Frankly, it's a little unclear what inspired the writing of this paper.
But perhaps they were just trying to make a splash with their title-- a little rhetorical excess is always forgivable. Their arguments against large-scale association studies are centered around two questions-- first, will most genetic variants involved in complex disease be detectable by association, and second, even if an association is observed, where will that eventually get us? Both of these questions are legitimate ones, yet I find their answers to both disappointing.
I. Will most genetic variants involved in complex disease be detectable by association?
For reasons I'm not going to explain, linkage studies generally have only the ability to detect rare alleles that greatly increase succeptibility to a disease. Association studies, on the other hand, have the ability to detect more common alleles that only moderately increase succeptibility. The authors note a number of successes in which an association study has identified such alleles. They then give a laundry list of problems with expanding these studies-- rare variants are important and missed by the association approach, gene-gene or gene-environment interactions as well, multiple testing is an issue when scanning the entire genome, genetic heterogeneity is likely when comparing populations, etc., etc., etc. All of these things are relevant, but honestly, this isn't news; these arguments have been around for quite a while.
Ultimately, after raising an issue, most scientists generally try to address it. These authors don't. The inital results from genome-wide association studies are promising, and it's perhaps ironic that, while they were busy writing this paper, other papers were published on clever new study designs and strategies for dealing with gene-gene interactions. The methods for ideally addressing the issues raised by the authors don't exist yet, it's true. But for some, that means there's still an opening to make a big discovery or develop a groundbreaking new tool. For the authors, it's almost like they don't see the point in doing human genetics research anymore. Indeed, their answer to the second question strengthens that suspicion.
II. Why bother finding genes that are involved in complex disease?
The goal behind finding genes involved in disease, any biotech CEO will tell you, is to gain insight into the disease mechanism and ultimately to disrupt it. In the wake of the sequencing of the human genome, companies were founded by the boatload to do just that. Most of them wasted away, as documented in a Slate article from 2005 subtitled "Why genetics is so far a bondoogle". But even that writer (presumably not a geneticist) had enough sense of perspective to realize that pronouncing the uselessness of genetics research after a few years of disappointments might be a little premature.
The authors of this paper aren't going to announce the death of genetics research either, but they don't seem to hold out much hope for it going anywhere. A comprehensive review of companies that have tried to make the path from genetic association to treatment would have been appropriate here. However, the authors give only one example-- their inability to figure out the role of the immune system in auto-immune disease. Genetic mapping has consistently associated the HLA region of the human genome, which codes for proteins important in the immune system, with diseases like Type I diabetes and Crohn's disease. But no one has really figured out how this genetic succeptibility works, or much less developed a treatment based on this knowledge. This is all true, but to generalize from trouble understanding the intricacies of the immune response to an impotency in all other aspects of understanding human biology is unwarranted.
If it were warrented, the entire discussion of the best approach to studying the genetic basis of human disease would be moot. It's obviously a waste of time to find disease genes if eventually they don't lead to any medical advances. The authors' point can't be that understanding the genetics of disease isn't worthwhile, otherwise they wouldn't have devoted pages to debating the best approach for doing so. Yet their sole example in this section makes the outlook appear somewhat bleak (again, I'm not sure what the goal was in writing this section). So I'll present a few counter-examples:
1. In the '70s, a devastating disease marked by early heart attacks and high cholesterol was mapped back to its causal locus-- a receptor for high density lipoproteins. This was an important insight into how cholesterol levels are controlled in the body, which eventually played a role in the development of the most popular cholesterol-lowering drugs on the market today-- the statins.
2. In 1960, cytogeneticists noticed a curious chromosomal abnormality in patients with a certain type of leukemia, which they dubbed the Philadlphia chromosome. As the decades went on and technology improved, closer and closer looks were taken at this chromosome, which was eventually shown to always be a translocation bringing together two genes, one of which was actived by this translocation. It was hypothesized that inactivating this gene could be a plausible treatment, and indeed, in 2001, Gleevac hit the market.
3. More recently, DeCode Genetics has turned an association between variants in a gene and risk of heart attack into a successful clinical trial. The drug isn't on the market yet, but it could prove to be the first big success of the post-genome era.
Yes, the successes I've noted have taken time (even decades) to go from a genetic association to a usable drug, but large-scale association studies promise to accelerate that process by allowing researchers to directly find relevant genes in an expedient fashion. Nothing happens overnight, even, believe it or not, solutions to the most prevalent diseases in Western society.
So the pessimism of these authors, while perhaps understandable, is unwarranted, nor does it make for a particularly good paper. I'm reminded of a another, perhaps more clever, paper with similar conclusions...ah yes, here it is:
National Science Foundation: Science Hard
 See here for a primer on the difference between linkage and association.