Friday, March 25, 2011

The Zen of GWAS: the sound of one hand clapping

So we've come to this: Nature is applauding the latest genomewide association study (GWAS) on schizophrenia as "welcome news" because it is "zeroing in on a gene" to explain this devastating disease whose etiology has been frustratingly elusive for so long (Hugh Piggins, "Zooming in on a Gene").  Many authors have found 'hits' by mapping, but most of them, if not perhaps all, have not been replicable.  The largest recent study we know of, prominently published (Nature, 2009), estimated that hundreds of genes contribute to schizophrenia.  Piggins does acknowledge that GWAS have been much criticized for explaining so little, but, he says, this one's different (well, at the very least, it'll sell more copies of Nature).

Speaking of copies, rare copy number variants (CNVs) have been found to be associated with schizophrenia and other neurodevelopmental disorders including autism. The operative word here being 'rare'. Copy number variants are generally large (1000 basepair or greater) genomic insertions or deletions, that, by definition, vary widely among individuals.  They're either inherited from a parent who carries the CNV, or arise anew. When CNVs were first recognized, it was thought that they would be found to be associated with many diseases, but the most common CNVs seem to not be disease-related at all.  After all, genomes evolve largely by segment duplication.
 
So given how Nature touts this result, we thought we must have misread, surely.  But, no, the paper confirms:
Here we performed a large two-stage genome-wide scan of rare CNVs and report the significant association of copy number gains at chromosome 7q36.3 with schizophrenia. Microduplications with variable breakpoints occurred within a 362-kilobase region and were detected in 29 of 8,290 (0.35%) patients versus 2 of 7,431 (0.03%) controls in the combined sample.
That's 0.35%, as in 3 schizophrenics per thousand.  That's a signal so weak that even a smoke alarm couldn't detect  it.  So, what's the real import of this finding?  Nothing new at all -- schizophrenia  is a complex disorder, or suite of disorders, that is multigenic, and/or has multiple different causes.  Like most other complex diseases, as has been shown over and over.

But the authors go on to discuss the gene (VIPR2) at the identified chromosomal locus that they think might be causative, and conclude, in what may be the Oversell of the Century to date:
The link between VIPR2 duplications and schizophrenia may have significant implications for the development of molecular diagnostics and treatments for this disorder. Genetic testing for duplications of the 7q36 region could enable the early detection of a subtype of patients characterized by overexpression of VIPR2. Significant potential also exists for the development of therapeutics targeting this receptor. For instance, a selective antagonist of the VPAC2 receptor could have therapeutic potential in patients who carry duplications of the VIPR2 region. Peptide derivatives and small molecules have been identified that are selective VPAC2 inhibitors, and these pharmacological studies offer potential leads in the development of new drugs. Although duplications of VIPR2 account for a small percentage of patients, the rapidly growing list of rare CNVs that are implicated in schizophrenia indicates that this psychiatric disorder is, in part, a constellation of multiple rare diseases. This knowledge, along with a growing interest in the development of drugs targeting rare disorders, provides an avenue for the development of new treatments for schizophrenia.
You may not have heard of this infamous gene, so for your edification, it's name is Vasoactive intestinal peptide receptor 2 (hence VIPR2).  The ultra high plausibility of this Major Gene for--what was it?  schizophrenia--is made clear by the sites in which it is expressed: the uterus, prostate, smooth muscle of the GI tract, seminal vescicles, blood vessels, and thymus.  Wiki adds as an afterthought that VIPR2 is also expressed in the cerebellum (whew!  A narrow escape chance for relevance?).

In fact, here's a section from GenePaint showing VIPR2 expression in a 14.5 day mouse embryo.  The gene is expressed where you see the darker blue -- the snout, the vertebrae, the ribs and lungs....  Not in the brain, but then this is only one stage in development, so it's relevance to brain function can't be ruled out.

But from the evidence, targeting this for therapy might ease digestion and calm the nerves.....including those in the genitals.  (So maybe schizophrenia is a sex problem, after all.)
 
We thought and thought what would be the right way to characterize this stunning discovery.  A supernova of genetics?  Darwin redux?  The sting of the VIPR2?  No, those images are too pedestrian.  We needed to go much deeper, to something with much more profound imagery, to capture what has just been announced.

Of course, it's possible that we've missed something in the story that is far more important than our impression has been.  It's always possible since we're no less fallible than the next person.

Nonetheless, based on our understanding of the story, we thought, well, Zen Buddhism is about as profound as it gets, in human thought and experience.  So we decided that the clamour of this new finding, the glory of GWAS, was the roaring sound of one hand clapping.  Listen very, very (very) carefully, and you, too, may be able to hear it!

10 comments:

  1. I like a good, snarky blog post as much as the next guy, but you’re way off base here. To begin with, it’s misleading to say that most of the “hits” found by geneticists have “not been replicable”. Since 2007, the vast majority of well-powered GWAS have indeed been replicated, across a range of common disorders. Furthermore, it makes no sense to compare the Vacic et al. study with a traditional GWAS in which a million or more common SNPs are genotyped and tested for association with a phenotype. In this case, the authors identified a rather small number of very rare number of copy number variants and tested for association. On top of that, you don’t make the critical distinction: common variants typically confer marginal risks that are quite small (odds ratio 1.2-1.4); in this case, the rare duplication on chromosome 7q36.3 is highly penetrant, with an odds ratio of 14. Even if that effect size is an overestimate, you’re still talking about a mutation that puts a carrier at very high relative risk compared to the rest of the population. You state that we already knew that schizophrenia is a complex, multigenic disorder. That’s correct, but why dismiss a study that actually goes to the trouble of identifying one of the genes? How is that “nothing new at all”, especially when VIPR2 seems to be such a powerful risk factor? And by the way, 0.35% may sound like a small number, but for a relatively common disorder like schizophrenia 0.35% probably adds up to a finding that is relevant to several thousand people struggling with this devastating illness in the United States alone. As for the expression pattern of the gene, can I suggest that Wikipedia is not the best place to get complete and accurate information of this sort? The paper itself makes it abundantly clear that the gene is actually expressed in several regions of the brain (suprachiasmatic nucleus, amygdala, hippocampus, hypothalamus), is associated with a signaling pathway (cAMP) that has been implicated in schizophrenia, and has established roles in learning and memory in model organisms. As such, the relevance of VIPR2 is clear. The fact that the duplication is associated with high risk of schizophrenia does indeed raise the possibility of intervention by reducing the activity of this signaling pathway, and the authors quite reasonably note the existence of selective antagonists of VIPR2 as promising leads in the development of a targeted therapy. In my view this is a good and important study, and doesn’t deserve the sort of glib takedown that I usually associated with blogs that are far less smart and stimulating than the Mermaid’s Tale.

    ReplyDelete
  2. First, thanks for the compliment you snuck in at the end of your criticism! Some of what you say I agree with (yes, it's good to identify the cause of devastating disorders when it can be done, e.g.). Perhaps our disagreement largely has to do with how we feel about the promise of GWAS and how to attack complex diseases in general. Perhaps you are seeing the glass as half full while we're seeing it as half empty.

    What first got us exercised about this paper was the idea that it proves that GWAS are doing the job. See our many previous posts on that subject. But, even if VIPR2 turns out to be a cause of schizophrenia, it can't be assumed that every case of schizophrenia will turn out to be due to a single gene. And, I would doubt that a pharmaceutical company is going to jump on the chance to develop drugs for a cause that accounts for 0.35% of any disease, though for the sake of those found to have this cause, I hope I am wrong.

    As for this gene being expressed in the brain, this is no surprise. Something like 80% of all genes are expressed in the brain.

    So, we don't argue with the possibility that a cause of some cases of schizophrenia might have been found, though we think the jury is still out. We argue with the idea that this study is another feather in the cap for those who believe that genomewide association studies are going to be a major step toward understanding complex diseases. We understand complex diseases -- it's likely that, by and large, major genetic findings for traits like these are behind us.

    ReplyDelete
  3. There is no doubt, as we've said many times, that the predominant view, and of course the control of resources (and publishing choices) favor the belief that this is stunning work.

    Many if not most genes are expressed in the brain so it's not surprising if the Viper gene is, too. But the list of circumstantial evidence that doesn't really bear fruit is long and, contrary to your assertion, replication is not the common finding--except for the really strong signals, and we've not said otherwise.

    However, even most 'strong' signals are relatively rare in the population. The argument that rare can still mean many victims is true, assuming the effects are not being overestimated, which is unlikely, and is a long-standing justification for ever-more, ever-larger GWAS.

    Focused studies following up reasonable GWAS hits have a history of, at least in some instances, showing the functional reason for the hit. But there is also a long history of insubstantial findings, too.

    The main reason for our reaction is the degree to which these findings are being heavily hyped, including by publishers and urged on by investigators lobbying for funding and anyone who thinks this isn't so is as oblivious as you suggest we are.

    If the finding is correct then follow up will show it and some day there will be treatment for some people. But it is at this stage not something that demands headlines, in my view.

    And there are many other relevant questions that we have asked about the epistemology and societal politics of GWAS studies that would apply here as well.

    But it's a free country, so you have every right to your view and, indeed, you are of course with the overwhelming majority.

    ReplyDelete
  4. Thanks very much for your replies. I appreciate the fact that you engage with your readers. I’d like to clarify a few points. First, as I tried to explain in my original comment, I don’t think the Vacic et al. study has anything at all to say about the power of GWAS to identify genetic risk factors for disease. The kinds of GWAS that you’ve been criticizing (sometimes rightly, in my view) survey millions of very common, low-penetrance single-nucleotide variants—markers that are only proxies for the actual functional variants that confer more or less risk. In contrast, Vacic et al. genotyped a much smaller number of rare, potentially highly penetrant copy number variants that are almost certainly functional mutations themselves rather than proxies. This is a very different experiment, and a very different model of genetic risk, which your post didn’t make clear (to be fair, the Nature News & Views article needlessly muddied the waters on this point as well). And of course I entirely agree with the statement that “it can’t be assumed that every case of schizophrenia will turn out to be due to a single gene”; in fact, I doubt that even VIPR2 duplications are sufficient to cause schizophrenia by themselves, despite the fact that they are probably the strongest single-gene risk factor identified to date (the odds ratio is 14, not 100). As for the issue of gene expression in the brain, I’m aware that most genes are expressed there. I only brought it up because your post had a big picture of a mouse embryo plastered in the middle of it whose only conceivable purpose was to imply that the gene had nothing whatsoever to do with the brain. Finally, you mention the issue of hype. Given the heterogeneity of causes underlying a complex disorder like schizophrenia, not to mention the complexity of the brain, it’s almost certainly going to be the case that all progress will be to some degree ‘incremental’. Does that mean that a journal like Nature should never publish a paper on this topic that falls short of reporting a convincing clinical trial? That would mean abandoning the field, and I don’t think that would serve either the field or Nature’s readership very well. (full disclosure: I used to be an editor at Nature Genetics).

    ReplyDelete
  5. Thanks. I agree that the more focused the more likely that what's found means at least something worth thinking about. Genomic mapping studies usually, almost necessarily, end up overestimating risk. And perhaps we were too affected by the commentary's claiming this as relevant to the GWAS idea.

    If one is thinking that a diagnosis at birth of positive for VIPR2 is meaningful, what does that mean in terms of absolute risk? And do cultural trends (even things like how much hyper-fast TV or video you watch, or things in diet, etc etc.) affect the risk currently being estimated?

    Does this mean that everybody is going to have to be followed up clinically for everything that the Collins-view of the world will end up diagnosing? Or everybody will be on a cocktail of maintenance meds for life? That is likely a goal of Pharma, isn't it?

    I guess I'd say that the claims of follow-up success that are made also should be judged, at least in principle, to include what else might have been achieved with the same funds. Even just antismoking campaigns or restrictions on SuperSizeMe's would be better translational use of resources. To me personally, it would be to spend funds on real applications to clearly genetic traits, like Huntington's, Muscular Dystrophy, etc., rather than the diseases, often luxury diseases, of the well-off when they get older (and I say that as one of them!).

    As to publishing, articles like that one ought, in my old-fashioned view, be in the Journal of Psychiatric Science (that is, a modest, un-hyped professional journal). With a journal like Nature, in it for the business and also (as with Science) for Big Status, it has the effect of systematically exaggerating findings that, as a number of people have pointed out, overstate not only the actual import of the finding (i.e., their true replicability), but the importance of the approach and of the findings application to society.

    Ironically, although I've been making points of this type, relevant to things at the time, for well over 20 years, and (I think, perhaps not modestly enough) that my assessments were roughly correct and for the right reasons.

    Perhaps my most noted major paper on this was with Joe Terwilliger in, yes, Nature Genetics. What we said would go almost unedited today, around 10 years later. I mention this because, ironically, also in that paper, was the story of calpain-10, touted as highly as anything these days (including VIPR2). But what happened to that as the 'diabetes gene'?

    And, one may ask, why haven't the countless other studies of schizophrenia found a good 'hit' in the CNV region? Perhaps in retrospect a blip in the manhattan plots might be detectable.

    Finally, and importantly to me, if VIPR2 is really even a modest contributing factor, given the clear evidence that many--probably hundreds--genes contributed to schizophrenia, each Viper-bitten person's risk will be different, because the major contributing genotype at these other loci, will differ. Not to mention their environments.

    You won't agree as an editor of a hungry weekly that this suggests quieting down these kinds of stories, and keeping where they belong (I think) in boring old speciality journals, until they prove to be really something worth international prominence.

    Anyway, again that's just my view, and I have no leverage on the system so it can do no harm!

    ReplyDelete
  6. Tx muchly, an informative and fascinating point-counterpoint discussion that is very illuminating to someone like me with only a modest comprehension of modern human genetics, and it's why I come back to this blog regularly.

    OK, back to your regular Nature-bashing program ;)

    ReplyDelete
  7. Thanks, OE. Glad you keep coming back, and glad to have your comments. (Didn't we bash some other journal once? Or even twice?)

    ReplyDelete
  8. It's not good for us to be making a career out of writing J'Accuse! posts, knowing that we are in the distinct minority. At least, we can say that we have been funded, and are funded, by the system (so this is not sour grapes), and we acknowledge that our lab is doing as much up to date genetics as we know how; at least we hope we're trying to come to grips with the real issues that we say should be studied (whether we're successful is entirely a different matter, and we make no claims).

    But if as it's said, one should never mix business with pleasure, in the sense of this discussion one shouldn't mix business with science.

    Nature Ltd is a business, and they have to sell product. So is Cell Press and Wiley and others. So what do you expect them to do? It's fair to say (as we do) that the AAAS should behave differently in regard to Science, but that is probably asking too much given the nature of our society.

    When journals decide not to publish negative results (good results that find nothing substantial are usually way more valuable than weak 'positive' findings), or to become known for only publishing one type of study (e.g., GWAS in Nature Genetics in recent years), that is the same as hyping a field of work--it's advocacy, not science, rather than letting the science itself show what's important.

    It encourages the marketeering of science that we're now experiencing to a huge extent, it discourages innovation (indeed, graduate students are trained NOT to be really innovative), and so on.

    Science, unlike television and magazines, is not about entertainment. It should, so to speak, be quiet, mostly boring, and done behind the scenes.....so that the truth can be pursued for its own merits, not for careerism.

    But we know that's not the current stage upon which we have to play.

    Nonetheless, every critique that could even slightly push things in a direction more likely to advance real, creative thinking is worth the effort.

    Even if, like the sound of one hand clapping, nobody will hear anything....

    ReplyDelete
  9. You gotta love this:
    http://xkcd.com/882/

    ReplyDelete
  10. Beautiful, Francesc! You know, Ken wanted to be a cartoonist when he was a kid. Maybe the time has come -- then we wouldn't have to write all these damn wordy posts!

    ReplyDelete