Personal genomics company 23andMe has always differentiated itself from its more sober competitors through an emphasis on collaborative, consumer-driven research - essentially, encouraging its customers to contribute their genetic and trait data to internal research projects designed to find new genetic associations. It is widely believed that generating novel associations between genetic variants and traits is actually the core business strategy for the company, although the precise mechanism for converting such associations into cash-flow remains unclear.
The company's initial attempt at obtaining detailed trait information was via a system of online surveys for customers designed to harvest information about traits ranging from left-handedness to dental health, cutely titled 23andWe. The major problem with this type of approach is that most typical personal genomics customers don't actually carry the types of juicy phenotypes most attractive for association studies (i.e. common diseases and disease-associated traits), a deficiency the company has since tried to overcome through targeted recruitment of specific groups of people for much larger association studies (thus far restricted to Parkinson's patients and pregnant women).
Today 23andMe launched a new initiative, the somewhat less cutely-titled Research Revolution, that seeks to cast the net much more widely in recruiting customers with specific diseases. The initiative was not widely publicised, but it warrants close attention - both for what it tells us about 23andMe's overall business strategy, and the lessons it might have for academic groups seeking to recruit patients for large-scale genetic studies.
How it works
Here's the summary. Basically, potential recruits can either sign up for a substantially cheaper Research Edition of 23andMe's service (which costs $99 vs $399 for the full service, but offers reduced content), or can simply sign up free if they are existing 23andMe customers. Both disease patients and healthy individuals - who will serve as controls in any subsequent association studies - then register their vote for their favourite common disease, from a list that is currently restricted to these conditions:
- amyotrophic lateral sclerosis
- celiac disease
- epilepsy
- lymphoma & leukemia
- migraines
- multiple sclerosis
- psoriasis
- rheumatoid arthritis
- severe food allergies
- testicular cancer
Any disease that gets the required number of patients (currently 1,000) will then be the subject of an association study performed by the company's scientists, using the genetic data from the assembled patients and controls.
It's worth taking a closer look at the reduction in content available from the Research Edition product compared to 23andMe's full service. While the cost saving is impressive (75%), you also lose several of the features that are probably of most actual value: for instance, results for all of the rare, severe disease mutations on 23andMe's custom chip (i.e. carrier testing), as well as all facets of ancestry testing. Perhaps most importantly for genomics enthusiasts, you also lose the ability to download your raw genetic data, preventing you from running your own analyses or using third-party software such as Promethease. In general I'd argue that new customers are still better off going for the full service than this stripped-down version, but I think the magical sub-$100 price tag will probably have a very positive effect on sales.
Will it generate novel associations?
Let me be perfectly frank - it's unlikely that a genome-wide association study with only 1,000 patients will reveal any novel genetic associations, especially for those diseases on the list that have already been subject to substantially larger, well-performed studies (e.g. psoriasis, rheumatoid arthritis, testicular cancer, celiac disease). There are also some serious doubts about the noise introduced by including individuals in a case-control study on the basis of self-reported disease status. In summary, I would be surprised if any of these early studies generate results that contribute in any meaningful way to genetic research.
So if this was simply aimed at recruiting 1,000 patients for each of 10 common diseases, it would be largely an exercise in futility. However, 23andMe's goals are clearly far beyond this: they aim to build stable, self-sustaining communities of potential research participants, that add new members over time and are available to add further trait data (for instance, to answer more detailed questions about disease progression) and also to engage in more sophisticated genetic tests (up to and including whole-genome sequencing, eventually).
Will this attempt work? I'm a believer. If 23andMe can recruit a solid core of patient advocates for a specific disease, those advocates then have a strong incentive to recruit further participants: increased numbers increases the probability of getting solid results for association studies done in their disease. As patient cohorts get ever larger and 23andMe's analysis becomes more sophisticated, there's every reason to expect that this approach will eventually yield the power required to generate novel associations.
If 23andMe's business model holds up - or, perhaps more to the point, if Google is happy to continue propping the company up - for the next few years, I wouldn't be shocked to see this research model eventually gather larger cohorts than even the largest academic consortiums, particularly for less common diseases with particularly strong grass-roots activists.
Implications for academic researchers
It's easy to be cynical about 23andMe's motives in launching this initiative - this is, after all, a profit-making exercise rather than a research charity - but I suspect that many academics will be watching 23andMe's progress in this venture with considerable interest. Modern genomics studies require mind-bogglingly large numbers of samples to achieve the power required to find subtle genetic associations, and recruiting those numbers of patients is far from easy. If 23andMe's model proves successful in recruiting and maintaining large communities of patients, I suspect large academic consortia will at least begin to consider the possibility of pursuing similar approaches.
That actually puts 23andMe in a potentially lucrative position. In order for academic consortia to pursue the 23andMe model, they need to be in a position to return comprehensive results from genome scans to their patients and controls. However, providing such complex information to a lay audience is extremely difficult, and probably beyond the means of most academic groups. That means (as I noted back in March) there's a potentially massive possible market for 23andMe here in providing a mediation service for returning research data to patients, and for providing the resources required to keep participants engaged actively in the research community.
But regardless of whether companies like 23andMe end up being physically involved in this mediation process, I'd suggest that academics need to take heed of the model the company is pursuing. It's likely that over the next few years the current model for returning research data to participants - i.e. don't - will become increasingly unpopular with potential research subjects, and indeed I'd argue that this model has always bordered on the unethical. Finding realistic ways of presenting large-scale genetic data to research participants is something that academic researchers will need to sort out soon, one way or another - and those that do it well, I suspect, will find it much easier to recruit and maintain their research cohorts.
- Log in to post comments
Daniel-
Thank you for a great review and summary, and for a brilliant last paragraph. So many ethical concerns can be eliminated or reduced by making subjects active participants in research. The old model is so entrenched, however, that I wonder if this "new" model will be adopted despite its numerous advantages and superior ethical treatment. On the other hand, who knows what will happen once people have access to their genome!
This announcement will undoubtedly result in more of the familiar criticism that 23andMe stands to strongly benefit from data obtained from paying customers, which they certainly might if the research is fruitful (indeed, let's hope it is fruitful!). However, I also know that if I were suffering from one of these diseases, I would have no problem with someone else benefiting from my DNA if something good came from it. Indeed, as a potential PGP participant, I'm already hoping that someone will benefit from my DNA even though I'm not suffering from one of the above-named afflictions.
Really a terrific, forward-thinking post.
Thanks Blaine. I suspect it will take a very long time for the old model to slip into the final grave it deserves - not just because it's entrenched, but also because of the non-trivial logistical problems it raises. It's likely that no one research consortium can afford to build the infrastructure required for appropriate return of test results to participants, so any research group that tried to do this alone would struggle.
There are two potential solutions. The first is to allocate money in initial funding applications that will pay for 23andMe et al. to return data to participants, which is likely to be unattractive to many academics (particularly clinicians) opposed to the whole ethos that 23andMe stands for. The second solution, which most be regarded as far more viable long-term, is for funding bodies such as the Wellcome Trust or NIH to pay for the construction of comprehensive, well-maintained databases and software for interpreting genome sequences that can then be employed by academic groups as needed - effectively, a public competitor to DTC genetic testing companies, with a mandate focused on the responsible provision of accurate genetic information to research participants.
Yes,
23andSergey would make a great clinical research organization. But as for as a Principal Investigator, still need questions answered.
Who? Who are the Scientists?
What? What exactly will they be looking at?
Why? Why is there no IRB? Isn't research on willing participants still research?
A good move to bring them away from the medical feel. But it changes nothing for now. Would have rather seen them be a recruitment partner for WTCCC or NIH....
That would add gravitas to them.
-Steve
Steve Murphy wrote:
The data access policies of both the WT and NIH mandate (if subjects' consent is adequate) that publicly-funded medical research data be made available to commercial entities as well as academic ones.
Therefore 23andMe will be able to add "their" (few) samples to downloaded GWAS data in order to perform a meta-analysis in a nice, non-reciprocal way. So they may get lucky, especially if they adopt the we're-not-really-looking-for-associations-any-more model of the The International Schizophrenia Consortium - actually I'd welcome a post on that paper!
I also agree with Steve, and perhaps disagree with Blaine on the continued ethical oversight, rather than gatekeeping, role of IRBs:
The IRBs function to make sure informed consent is adequate, investigators credible, studies are well-powered etc. some of which cannot be left to caveat emptor ...
It is an interesting move, you're probably right, the 1000 is just for starters, but could there be some possibility of interesting findings if they restricted the search to a few of the candidate genes for each disease? Plus they may be able to get more detail on the environmental exposures of each subject (diet, exercise, etc, which is largely ignored in GWAS so far)and may be able to get stronger links with gene + environment + disease?
The model is not a new one but is one made possible through the genetic data - in our Eurogene project we are working supporting systems for similar types of data and patient recruitment - NOT DTC though but through clinics (come along to our workshop, it's part of NuGo Week, 31Aug - 3 Sept, in Tuscany, Italy, http://www.nugo.org/everyone/39927)
Keith
The IRBs function to make sure informed consent is adequate, investigators credible, studies are well-powered [emphasis mine]
mission creep?
Steve and Neil,
Interesting point about the IRB oversight of things like informed consent. I would rather see a commercial entity like 23andMe partnering with another research entity (such as a university) rather than enact new legislation, but of course informed consent is vital to ANY research project.
With my statement I was referring to the current model where once a participant fills out a survey and submits blood or other sample, their interaction with the research project is finished. Even if the subject understands that she will not receive any results after reading IRB-approved consent, I completely agree with Daniel that making no effort to return research data to subjects âhas always bordered on the unethical.â This is especially true if the research involves a condition that can be treated by medicine or lifestyle changes now or during the life of the subject. Of course this will add cost to any research, but researchers already accept many other costs in order to be as ethical as possible (the IRB approval process, for example, and confidentiality procedures). Having subjects who actively participate in research and receive their individual research results will eliminate many of these ethical concerns.
On the principle that there will be about one participant born per minute, they will surely find the recruits that they require for a nice publicity stunt. Very little else will be achieved apart from some unreliable studies of dubious ethical provenance that may bring genetics back to the low esteem it held in the 1930's. Self-funded Eugenics anyone? How about a dating site for the Master Race? Go see your doctor if you're not well or worried about your future health, join a real patient support group (there are plenty) if you want to help on a broader front.
> Self-funded Eugenics anyone?
Yes, please. It's called liberal eugenics.
Hi Neil,
...the we're-not-really-looking-for-associations-any-more model of the The International Schizophrenia Consortium - actually I'd welcome a post on that paper!
You mean the "look Mum, we got a Nature paper using a P value threshold of 0.5" paper? I've been trying to eke out an hour or two to blog about that all week - it sets some very strange precedents for the future of GWAS...
You mean the "look Mum, we got a Nature paper using a P value threshold of 0.5" paper? I've been trying to eke out an hour or two to blog about that all week - it sets some very strange precedents for the future of GWAS.
Yep, that's the one.
I noticed:
http://wiringthebrain.blogspot.com/2009/07/hot-news-in-genetics-of-schi…
but struggled to work out what was special about schizophrenia's "polygenic" model that means it is not the same "multiple genetic and environmental risk factors" that the rest of us use.
The author of this blog concludes:
Personally, I would have thought the "rare variant" hypothesis is unnecessary at this point: that [a] at least some psychiatric diagnoses are social constructs so that "schizophrenia" may be a raft of different disorders, and [b] heritability need not be genetic - shared upbringing must take a role. I'm thinking Larkin here ...
I can't figure out who would sign up for this. I've been wanting for years to have enough money to pay for this kind of study, or for the prices to drop. I went and looked at what you get for your $99 and--nothing. You get absolutely none of the services that anybody is actually interested in from 23andme (the ancestry and disease screening parts--I'm only interested in the ancestry aspect, honestly). So you're just paying to be part of their study. I support research, but no thanks!
Paying to have no control over what is done with your DNA and who has access to it. Sounds like a bargain to me...