The Economist has a rather distressingly bad article by the evolutionary psychologist Geoffrey Miller, about the supposed general failure in human disease genetics over the last 5 years. The thesis is that Genomes Wide Association Studies (GWAS) for common diseases have been a failure that geneticists are trying to keep hidden, and that the new techniques required to solve the problem of disease genetics will raise ‘politically awkward and morally perplexing facts’ about the different traits and evolutionary histories of races. The former claim is pretty much the same as Steve Jones Telegraph article earlier this year, and is just as specious. I will look at both claims separately.
A quick point of terminology: Miller uses ‘GWAS’ to refer to studies that look for disease association in common variants using a genotyping chip, and acts as if sequencing studies are not, in fact, GWAS. In fact, a sequencing association study is just another type of GWAS, just looking at a larger set of variants.
The Death of GWAS
A quick admission; I am too young to have experienced what discussions were underway five years ago, when scientists were planning the first GWAS. I was 18 when the WTCCC was founded. It is possible that every scientist involved said “lets set out on a project to perfectly account for all of human variant in disease” (though I saw a talk by Vincent Plagnol in 2006 that directly contradicted this). If you wanted to account for all variation in susceptibility for disease, then the current situation, in which we can explain less than about 30% of variation, will be disappointing.
However, if they did say that, they were being pretty stupid. For disease researchers, accounting for all the genetic risk in a largely non-genetic complex disease isn’t a hugely useful endeavor. What we care about is finding the mechanisms by which diseases start, the biological pathways that are disrupted, so we can figure out treatments. And, by these standards, GWAS have been a huge success. In some diseases, such as Inflammatory Bowel Disease, entirely new and completely unexpected pathways have been discovered (such as the Autophagy pathway), as well as giving new information on previously poorly understood cell types like Th17. In other diseases, new insights into existing pathways have been given; the autoimmune problems that lead to Type I Diabetes are being cracked wide open.
My interest is in autoimmune disorders, so these are the systems I know about in detail. But I have looked into obesity and Type II diabetes, and a similar story can be told there.
Miller spins a strange story that will be unfamiliar to disease gene researchers. He makes lots of odd mistakes, like saying that we have focused only on coding regions, which is entirely false; in fact, a large proportion of GWAS-discovered variants are nowhere near genes, and are telling us something interesting about non-coding genetic mechanisms. He says that we have a ‘couple of hundred’ GWAS hits that do not replicate across studies, which is not true; we have ‘a couple of hundred’ well-replicated and rediscovered GWAS variants, and thousands of non-replicated variants. Likewise, contrary to his implications, no-one keeps the ‘missing heritability’ problem hidden, and if there are ‘small, discrete conferences’ to talk about the problem, I have never met anyone who was invited to one (though the question was discussed at ASHG 2009, hardly small or discrete).
GWAS and Population Evolution
The final part of the article, about race and evolution, is just an attempt by the author to shoe-horn one of their own interests into a unrelated subject. He believes new sequencing GWAS will start confirming the many currently ill-evidenced hypotheses put forward by Gregory Cochran et al, about racial differences in various traits being explained by recent selection on genetic variants.
Sequencing GWAS are not the right place to look at these sort of population evolution questions; GWAS have to restrict the genetic diversity in their study groups, so as to keep down noise and reduce artifacts. As I discussed in one of my ASHG posts, if fact we will have to be even more careful to restrict genetic diversity in sequencing GWAS, making these studies even less useful for answering questions about ‘race evolution’ than current generation GWAS.
In fact, there are many datasets available now that are perfect for answering questions about the evolutionary histories of different populations. For instance, the HapMap project contains genotype data across 11 human populations, and the 1000 Genomes project is extending that data to a greater resolution. Pardis Sabeti is pretty well-loved in the genetics community for her work using these datasets to track down evolutionary pressures, and Chris Tyler-Smith and his group are doing similar things.
We are already revealing these ‘politically awkward’ results about human evolution, and they are these: there have been a few hundred detectable signals of strong, diverse selective pressures over the last tens of thousands of years. The dominant selective pressures in Europeans have had for skin color and lactose tolerance; in Africa it is the immune system; in East Asian populations it is sensory perception. All populations are under selection for metabolism. We know these things already; they haven’t created any massive political stirring, and GWAS results aren’t going to add anything new to this, with the possible exception of telling us which genes confer diabetes risk in India but not Europe, and other similar results.