The new study, by Neil Risch and Kathleen Merikangas and others (by no means skeptics about the possibility of explaining complex disease with genetics, though they have cautioned in the past that gene by environment interaction is important), is a 'meta-analysis' of 14 studies that attempted to replicate the original finding, but were less successful. Risch, Merikangas et al. reanalyzed the data and did not find an association between the serotonin gene and depression, though they confirm that life events are significantly correlated with this mood disorder.
Regular readers won't be surprised to hear that we aren't surprised by these results. We are interested, though, in the first two sentences of the JAMA paper:
The successful statistical identification and independent replication of numerous genetic markers in association studies have confirmed the utility of the genome-wide approach for the detection of genetic markers for complex disorders. However, recent genome-wide association studies have also indicated that most common genetic risks, at least when studied individually, are modest in magnitude, with relative risks in the range of 1.3 or less.
It's almost required these days for a genetics paper to start out by proclaiming the success of GWAS--so like most geneticists, these authors are in the GWAS camp....but then they aren't. Indeed, it's now becoming fashionable to want to have it both ways: GWAS have been a great success, but actually they haven't, so we need a lot more money to do other things--like whole genome sequencing on everyone. And that, in essence, is another form of GWAS but on a grander scale because one still must make statistical associations between variants and disease phenotypes.
The individual sequences of Watson and Venter that are on display (with others in the pipeline) already show thousands of previously unknown protein-changing variants, plus additional thousands of 'novel' SNPs of unknown function. So the push for huge studies of this type are still based on technophilic wing-and-prayer promises to a great extent. But nobody is willing to say: pull the plug on these approaches and try something more likely to identify meaningful causes, including meaningful genetic causes, of complex traits. Too many vested interests are at stake.
In response to an article we published a few years ago in the Int. J. of Epidemology (2006 Jun;35(3):562-71), Merikangas basically said that discovery was serendipitous and even if our skepticism was justified the money should keep flowing because, eventually, something would be found. This is not a novel argument and indeed goes back to the basic Baconian idea of induction: keep observing and the theory will emerge from the data. To us in this current context it's an ultimate form of self-interested last resort.
But, this discussion is a road we've traveled down a number of times already in this blog. This new paper certainly won't deter researchers from continuing to search for genes 'for' complex diseases like depression, schizophrenia or autism and neither will we. Biologically, these traits may not be particularly different from physical traits like obesity, diabetes, or cancer. But behavioral traits are different in two very important ways that are relevant to issues of science policy: First, they are exceedingly susceptible to cultural environmental experience and effects. Even when these interact with genetic variation, it is the cultural factors that are not only clearly preponderant but also most malleable. Secondly, they are socially sensitive in the sense of potential for real abuse. We can't forget history, which is rife with arguments about biological inherency that are used to discriminate against classes of people, be they nationalities or 'races'.
At some point, surely even the most dedicated gene hunters will acknowledge that this approach isn't working for complex disease, and will begin to rethink the problem. Network-based thinking, that is, treating systems of interacting genes as wholes, could be a way out, if it has to do with therapy. But probably not if it means simply identifying every variant each person may have in the countless genes in relevant networks.