In October last year I reported on a presentation by direct-to-consumer genetic testing company 23andMe at the American Society of Human Genetics meeting in Honolulu, in which the company described results of genetic association studies performed using combined genetic and survey data from their customers. The results of their study include replication of several known associations for traits like hair colour, eye colour and freckling, as well as the discovery of previously unpublished associations for things like asparagus anosmia (the ability to smell urinary breakdown products after eating asparagus) and photic sneeze (the tendency to sneeze when entering bright light).
This research has finally been published in PLoS Genetics, one full year after the manuscript was originally submitted to the journal on June 22, 2009. In a separate article, PLoS editors Greg Gibson and Gregory Copenhaver explain that this delay was due to six months of investigation into the issues of ethical review, participant consent and data access in the context of 23andMe's research. As a result of this process, 23andMe announced yesterday that all of its research will now be conducted under the auspices of a formal IRB.
In a post on her blog, 23andMe co-founder Linda Avey describes the publication as "historic", and I think this is no exaggeration. 23andMe has gathered an unique, actively engaged participant base of customers willing to contribute their data to the company's research efforts, many of whom (myself included) are excited to see their information being used to identify entirely novel scientific findings.
While it's easy to scoff at the traits assessed in this paper, the company is beginning to assemble a research base with growing power: the company tells us that of its 50,000 customers, 29,000 have participated in research surveys, enabling over 650 simultaneous genome-wide association studies. And because many of the company's customers are actively engaged, there's room for promising associations to be chased up with more detailed surveys and longitudinal studies. The power and flexibility of this approach would be the envy of many an academic researcher.
In a brilliant article on Google's Sergey Brin this week in Wired, Thomas Goetz illustrates the utility of the 23andMe approach with a simple example: a massive study of the association between glucocerebrosidase mutations and Parkinson's disease published late last year, the key finding of which was successfully replicated by analysis done by 23andMe on their Parkinson's recruits.
After a month of bad news for the personal genomics industry, it's good to be able to report something positive, and this certainly is positive: it's a demonstration of what can be achieved when research participants are given an opportunity to engage in research on their own data. I can only hope that academic research groups are paying attention, and thinking about ways they can leverage the same effects.
The mail in kits is what gives this such great flexibility. Perhaps a team doing the same yet giving away kits for free and with strategies to avoid Hawthorne effect will springboard this effort. Because, after all Academia is not a Corporation, yet.
The ongoing attempt by governments and pro-government activists to exert control over individual access to our genomes may make mail order *anything* very difficult to do in the future.
Moreover, anyone who reads this article (http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.10009…) must be struck by the level of bureaucracy built into the editorial and academic process. Drs. Greg Gibson and Gregory Copenhaver have inadvertently shined a spotlight on the realities of academic politics.
One notes immediately the aspersions cast upon the ethics of the eeeevul megacorp:
In this case, concerns were raised that a commercial IRB, paid for their opinion by the company, is not in a position of independence, but this is standard practice in the pharmaceutical and biotechnical industries, and similar concerns can be raised over the independence of University boards considering multi-millionâdollar studies
One is amazed that the similar concerns were even acknowledged in a piece like this, grudging though the acknowledgment may be.
the public good of disseminating valuable science performed by commercial entities
Note the implicit contrast in those words! Of course valuable science is not usually performed by commercial entities. Only a fool would think that Intel might be pushing the boundaries of semiconductor physics, Google advancing machine learning, or Roche innovating in synthetic chemistry.
Those corporate plutocrats are only publishing things that would advance their business and make a profit. You know, profit -- selling something for more than it costs you to make it. That's wrong of course. After all, if Intel is charging Joe $100 for a processor that costs them $10 to make, it must be wrong because Joe could have made that same processor in my backyard for $9!
(That's the deep comparative advantage argument for profit, by the way, folks.)
Unlike corporations, academics are pure of heart. They would never publish something that advanced their career. They would never rubbish a competitor's paper out of spite and never share confidential papers before review. They would never abuse their graduate students or suppress politically inconvenient research.
The PLoS journals are unfortunately the worst in terms of this kind of self-important anti-commercial attitude. Academics are assumed to be above reproach and businesses are guilty until proven innocent. The problem with PLoS in particular is that they confuse their competitors (the for-profit journals) with for-profit entities *in general* and assume them to be evil.
This is going to leave them very vulnerable to a journal which is to PLoS what "open source" is to "free software" -- taking the good aspects of the openness model without the anti-business baggage.
The editors of PLoS Genetics recognize that the decision to publish this study, without IRB review as human subjects research and with some concerns over the consent document, and the fact that there is limited access to the raw data, will not sit well with some, perhaps many, readers.
What a JOKE. How many papers are published in PLoS that include the raw data? You can bet right now without looking that 80% or more of the articles published in the last month of the PLoS journals do not make all the code and data available. In fact, you can bet that Gibson and Copenhaver have not made all their published data available.
@Tina Turner you make some good points. The commercial conflict of interest exists of course and should not be ignored. However ALL authors have a conflict of interest, academics especially so in this era of having to publish vast quantities of papers to get grants, tenure, promotion, or whatever. History shows that most scientific fraud has been conducted by academics. The days of academia vs. industry as two opposite poles where traffic was one-way from the first to the second are over, they really finished in the 1980's
It's all about statistical power.
All of the "unfulfilled promise" of the Human Genome Project (should we have spent that few billion on wars instead?) will be fulfilled as we get more and more statistical power.
It's like a person looking at a 50s era computer calculating 435*786 and saying, "I could do that with paper and pencil!"
Some people just like to be sticks in the mud.
Another good news for GWAS:
The study used a restricted cohort of people adjudged to be Northern or Western European.
So the findings are really only applicable to those people, and not really of much use to anyone else.
One of my criticisms of 23andMe has been its bias towards European (Americans) of North or Western European origins. In other words, SW, SE Europeans, Middle Easterners, South Asians, SE and NE Asians, Australian Aborigines, Amerindians, Pacific Islanders need not apply, go elsewhere as you are not wanted. WASPs only.
Ponto... you are incorrect. If anything they are not bias towards Europeans. I came up as being ~10% Asian and their results have been more in line with the America Indians than with Europeans (the reasons are complicated). Minus this, I actually agree with the approach to race selection. If a gene is found in 80% of Europeans and 20% of Asians, they'll associate that gene with Europe, where other companies will say you're 80% European and 20% Asian. Their analysis isn't perfect, but ancestry is not an exact science.
Somehow I just can't care about this. Genomics is moving into a post-GWAS world -- or at least a post-common-SNP-GWAS world, because those are proving to have very little predictive power, and are useful mostly for preliminary pathway highlighting.
Until they're using far more high-res scanning than is now commercially feasible, this feels like getting excited about a circa-1996 StarTAC when everyone's getting an iPhone.
Well, there's still things that can be done with GWAS data. For instance, I suspect we'll see some use of imputation (using reference data from sequenced individual to infer genotypes at untyped positions in people's genomic data) - this is by no means perfect, but it will add a little extra predictive power.
But yeah, the future is sequencing. But that won't invalidate 23andMe's model: they've collected phenotype data from a group of fairly engaged research participants who will be first in line when they launch their sequencing product, and that will start to provide some seriously powerful research fodder.
That's assuming the FDA actually lets them launch a sequencing product, of course...