I agree: GWAS is of limited value and this probably contributed to deCODE's demise. But whatever deCODE's fate, if whole human genomes can be sequenced for < $2000, isn't it about time we stopped kicking GWAS's ever-stiffening corpse? Second, just because something is not a medical necessity, does it follow that it is worthless?
if whole human genomes can be sequenced for $2000, isn't it about time we stopped kicking GWAS's ever-stiffening corpse?
Does she just mean hey-- let's forget about it it's old news? Well on the one hand there are still plenty of me too GWAS studies appearing last time I looked.
On the other hand perhaps our experience of GWAS should inform how we continue with genome sequencing.
Second, just because something is not a medical necessity, does it follow that it is worthless?
No, but no-one is making this strong statement. Rather they are making a statement questioning it's cost benefit ratio (i.e. high cost little benefit). My own feelings are that the initial GWAS studies were important and pioneering but once it became clear that they seldom uncovered key or medically useful heritable traits, then its use should have diminished as the costs should have been seen to outweigh the benefits of spending the money on other approaches.
Clearly however Nature Genetics feel that if something was expensive, high-tech- and time consuming then it belongs in their journal regardless of the interest of the result.
Regarding WGS its clear that this is very powerful for finding high penetrance mutations at relatively low cost. Pioneering studies will no doubt tell us how good it is at finding low or medium penetrance variants from large banks of patients. Once we know we should sit down and question whether we have got value for money before pursuing this willy nilly for the next 5 years.
First, a slight correction: Now that I've had my genome sequenced, I can say definitively that I do indeed have a Y chromosome, however humble and shrinking it may be (I'd hate to have to change my blog to "GenomeGirl" at this stage of the game).
1) I think it's important to distinguish between GWAS as research and GWAS as a means to make solid predictions about traits in the absence of other information. I'm saying that to keep belittling GWAS for its failure to deliver the latter has, IMHO, become tedious. Maybe I read the New England Journal too much.
2) "No one is making this strong statement." I would need more than two hands to count the number of clinical geneticists I've interviewed for my book, seminars and meetings I've attended, and screeds I've read that either explicitly say or insinuate that, for the moment, personal genomics is essentially worthless. I'd be happy to send you a reading list.
3) "Value for money"...Prior to the launch of the Human Genome Project, the critics complained that the HGP would be too expensive and deprive small science (aka "real science") of scarce resources. Today, despite genuine disappointment in genome-based drug development, diagnostics, GWAS etc., I'm hard-pressed to find anyone complaining about the $3 billion we spent on the HGP. I'd say that at $2000 a pop, pursuing WGS "willy nilly" is exactly what we should be doing.
The problem with GWA studies is that they have been done by geneticists with very little account of epidemiological principles eg misclassification, confounding and the effects of incomplete penetrance. To collect thousands of cases and believe that these represent the same gene effects is far too simplistic. The cases should be subclassified based on biochemical, histological or physiological grounds - not all breast cancer is the same disease and different genes will be relevant. To then gather controls (ie someone not affected) in the belief this will represent those not genetically disposed to the disease in question is too stupid for words. None of these studies have considered the role of environment on gene penetrance. This means the controls are a mix of peolple who might some day get the disease in question or someone with susceptibility genes who has not had sufficient environmental exposure for their genotype to express the illness. The controls in these studies are poorly selected, poorly defined and full of potential cases. Defining cases based on a "clinical diagnosis" is according doctors far more in sight into disease pathophysiology than they deserve. Until GWA studies are better designed using better phenotypes, then they will fall far short of their potential.
Apologies for misinterpreting your name.
1) Despite your clarification I'm confused as to your meaning. GWAS means 'genome wide association study'-- which take at least hundreds and more likely thousands of cases. Personal genomics seems to be what you mean.
2) I'm unconcerned by people who want to have a SNP chip done for them. Personally I see it as an interesting medical curiosity but -- except in very rare cases--you would get far more predictive info from a few simple blood tests.
3) The human genome project is nothing like GWAS. The first was a pioneering project that invented all our modern sequencing technologies, made possible all high throughput techniques and produced a blueprint for us to compare all personal genomics against. Likewise the HapMap project, the ENCODE project, and even the 1000 genomes project are all important pioneering attempts to understand variation and meaning within the genome. GWAS for weak heritable traits is just flogging a dead horse. My fear is that even if WGS proves equally poor at finding the 'missing heritability' we will continue to flog the dead horse -- but with a new whip.
So your argument is that the major problems with GWAS are:
1. Incomplete penetrance of disease risk variants;
2. Heterogeneity of disease phenotype in cases;
3. Presence of unknown disease cases in control cohorts; and
4. Gene by environment interactions.
Every single one of these issues is discussed at length in the GWAS literature. In some cases (e.g. incomplete penetrance) the issue is explicitly accounted for in the analysis of GWAS data, while in others (e.g. presence of cryptic disease cases in the control cohort) the magnitude of the problem has been modelled and shown to have relatively small effects on power for most diseases. Disease heterogeneity is being tackled by increasingly more sophisticated endophenotype analyses. Gene by environment interactions will be extremely difficult to tease out, but a lot of ink has been spilled on this issue in the literature, and very large longitudinal cohort studies will provide data to help start teasing these things out.
There are plenty of problems facing GWAS, just as there are for any other scientific technique. However, the notion that these problems are being ignored by the researchers doing the studies is totally false.
GWAS for weak heritable traits is just flogging a dead horse.
The heritability of a trait has little to do with the underlying genetic architecture. Recall that human height has a heritability of ~80%, but it influenced by many loci of small effect, while (for example) macular degeneration has lower heritability (see twin study here--50% concordance for MZ twins, and 25% concordance for DZ twins) but is influenced by at least one locus of large effect.
It a trait is heritable, the only way to see what the genetics underlying it is like is to look (more or less).
1. In spite of the tremendous efforts carried out by the scentific and very clever medical communities during centuries, the pathogenesis of complex diseases is mostly unknown.
2. GWAS is a consequence, and essentially derived, from the Human genome project. To date is the most powerful (and unbiased) strategy to pick up genes for complex diseases.
3. GWAS is based on epidemiological and statistical methods and theories (the same technologies employed to discover the relationships between smoking and lung cancer)
4. We are facing with a huge amount of data and variables and we are looking for the best methods to organize and integrate the information. The idea is to improve the knowledge of many complex traits. This is the true challenge irrespective you are employing hundred thousand variables (GWAS) or complete genomes (3 billion variables per individual).
5. Most of GWAS researchers are working into a multidisciplinary teams ie integrating clinical, molecular, epidemiology, statistical, informatics, mathematics researchers.
6. Genetics research is not panacea for medicine is just a novel technology that HAVE TO HELP US to improve the management of disease and health states.
7. So we are not flogging a dead horse, we are just applying new technologies (as usual in medical research) to discover whether or not that new methods can help us to prevent, diagnose and treat medical problems without any solutions to date.
GWAS was originally used to discover genes that are involved in complex diseases. The facts show that it does this job superbly, over 200 new loci were "discovered" in the last few years compared to less than a hundred discovered in the last 30 years. This gives a lot of new targets for researcher to conduct research about how these new loci are functionally linked to the diseases.
We should expect that as the sample size of GWAS increases, we will have the statistical power to detect rarer variants that can cause diseases. I believe that then more heritability can then be explained.
But if you are talking about applying GWAS results to personal genomics, then that's another story. But this is not what GWAS was set out to do in the first place.
In regards to the above comments, which company offers a $2,000 fully sequenced human genome?
I really, really like the Single Molecule Sequencing approach (http://www.helicosbio.com/), given how important micro RNAs are to protein expression. When might this technology be available for the masses (the $1,000 personal genome)?
I have A LOT of relatives with cancer and severe cardiovascular disease, despite a healthy lifestyle... after a lot of genetics courses including human pedigrees, I am anxious to know my genome.
The $2000 figure is misleading - Complete Genomics recently reported sequencing a human genome at this cost, but the figure only includes reagent costs. The cheapest retail whole genome sequence currently available is $48,000 from Illumina, but it's likely that sequencing will become available for well below $10,000 in 2010.
I would concede that if we're talking about a whole euchromatic human genome sequence then yes, $2000 is still misleading. But if Knome is doing exomes for $24k, I expect that DIYbio types will be doing their own exomes within a year. I don't think anyone doubts Daniel's prediction.
The other point that bears making is that if we're talking about finding disease susceptibility genes, a whole genome is probably gratuitous in most cases. I would bet dollars to donuts that if you asked someone working on your favorite common disease what genomic data would be most useful to have from patients, she would be more likely to want sequence from a list of 50 or 100 or 1000 genes than a complete diploid sequence at high coverage.
We're all becoming cyborgs anyway. Rant or no.
Thanks to Misha and Daniel for these great posts.
I have to point out to Stephen (who apologized for "misinterpreting" Misha's name) that the name Misha is an extremely common male nickname for Michael or Mikhail in Russia and elsewhere - similar to Mike here in the US. It helps to pay attention.
A friend told me about this link. Looks like deCODE is not fully dead by now.
www decodeme com / data-upload
What are the lessons we can learn from deCODEme's demise?