This company bothers me immensely. They take your $1000 and a buccal swab, and run it against an 550K SNP chip (beadarray, actually). That's all. The data which is generated effectively tells you what base your cellular DNA has at points along the genome where humans are highly divergent (these points, SNPs for Single Nucleotide Polymorphisms, tend to be passed from generation to generation in blocks which are resistant to breaking up during recombination; therefore, they reveal a lot about a person's ancestry, and the combinations of blocks present can predict some types of disease risk). It sounds cool; I'm not so sure that it's the best place to spend $1000 at this point in time, but hey -- SHINEY!!1 People will probably do what they're told and burn some cash on it.
This is all well and good, I suppose. You have a service of fairly dubious utility being offered, to a bunch of people who likely overestimate its utility. In other words, it's just like any other service in that respect. Maybe less useful than most. (edited to remove erroneous Affy/Illumina references -- with that sort of an error, you better not regard any of the rest of this post as definitive either!)
What sucks, in my mind, about this situation is that the company doesn't seem to be driving any real innovation. Right now, you can have your genome run against a dense panel of SNP markers, and find out what it contains at swab time. That's non-exhaustive, but statistically it barely matters because if you have major new mutations in conserved genes, you're probably dead. Most such mutations are fatal. The ones that aren't continue to be passed around as time goes on. But there are a number of other influences on the way your body ages, gets diseases (or doesn't), passes diseases on (or doesn't), and so on, which we DO know about. First off is epigenetic influences such as whether some of the cytosine bases are decorated with methyl groups. If 23andMe has any intention of offering new scans as the methylation chips become more useful, they're not telling. It will be some time before the success of genome-wide association scans, such as the successful efforts to locate SNPs and SNP (haplotype) blocks which put a person and their offspring at high risk for Crohn's colitis, is wide-spread enough that the genetic information is useful to the average person. The epigenetic and environmental information carried in a person's cells, and the interaction effects of all 3? Could be decades in the most optimistic case. If you think brute-force computational power will solve that, let me know how your protein folding programs are doing. Brute force CPU power was supposed to solve that, too.
Meanwhile, companies like Helicos and Pacific Biosystems are producing technology to cheaply and completely resequence a genome using whole-genome amplification from small samples. They're getting smacked around by the bear market in biotech, but if they succeed, and it looks like they will, then instead of a panel of 1000K SNPs, you'd get ALL of the information in a snapshot of your genome. All 3 billion bases, maternal and paternally derived copies -- remember, you can have different maternal and paternal alleles -- which means all the repeats, microsatellites, and other bits that the SNP chips don't get. (There are plenty of other gotchas as well -- but this is getting really long and I'm not in a mood to edit)
About the use of the word ''snapshot'' -- it's because cellular DNA appears to change in response to cellular RNA manipulation. Tom Cech won the Nobel prize for showing that RNA by itself could store, manipulate, and transmit information at a cellular level. More and more, it's appearing that RNA is actually running the show even in the presence of DNA, and DNA is more like short-term storage for the transient instructions from the RNA. So the DNA, rather than being static, is not just getting methylated and wrapped into bundles, but literally changed on the fly. (If the preceding story is correct -- which no one seems to know)
So what they're assembling at 23andMe is a big tissue bank, a collection of mailed-in buccal swabs, and they're offering the (IMHO, dubious) service of guiding you through that information. There are some useful things you can learn from your genome even now. If you and your spouse have certain mutations -- say, you're both carriers for Tay-Sachs -- then you know right off the bat that 25% of your children are likely to have the disease. If you have certain variations in, for example, TCF7L2 (see http://www.nature.com/nature/journal/v445/n7130/full/nature0...) your risk of type II diabetes is increased (though this is not nearly as conclusive as studies such as those for Crohn's). But most diseases aren't as simple genetically as Tay-Sachs or cystic fibrosis. Most are complex, apparently the product of environmental, genetic, and ''other'' influences. So what you are NOT going to find out definitively is what treatments you might respond to, WHETHER you are more likely to die of colon cancer than anything else... not solely from 23andMe. Not at this point, when there is a very great deal of uncertainty even within the statistical genetics community as to the overall utility of many of these expensive genome-wide association studies. Are they producing results? Fuck yeah! Are those results directly translating into cures or treatments? Uh...
If the current perception is correct, and RNA + environmental/epigenetic influences are in fact changing the genome on-the-fly, then a tissue bank that can re-run the snapshot you sent in (as a buccal, aka cheek, swap) against newer technologies may be of limited utility. The inherent utility is, in that scenario, delivered by the newer technology, and the expertise to run a 'hot' scan on affected, unaffected, or potentially-affected individuals at a particular point in time.
It's unclear to me how their infrastructure could help with this final item, and for all of Brin's wealth, I still don't see them bringing humanity closer to that point. The information could hurt you, by making you uninsurable, but I can't see a snapshot of potential risks helping people the way that newer technologies offering time-lapse snapshots of a person's genome may. That's why it bugs me.
I'm not sure your anger is well placed. 23andme gets an undue amount of attention so it's natural to feel frustrated by the media hype. But 23andme aren't (at this stage) lying to customers about what it does, and I'm pretty sure most customers will be interested in the knowledge not simply the utility. Also ruling out strongly genetic diseases is probably quite interesting to a lot of people, regardless of how useful it is in a public health sense.
Most likely their initial scan procedure isn't of great utility but does set up a market for people based on their genetics. A whole realm of industries could spring up which could be very interesting. I don't know why you think its 23andme's job to carry the entirety of biotech research, but if no one can make any money from personal genetics until we know everything there is to know about the genome then I doubt we'll advance the field at a great pace.
It's not their responsibility to do much of anything. But they're uniquely well positioned to push a lot of these things along, and it's a bit disappointing that they opt not to.
Plus, selling genomic information from your customers, in aggregate or not, is a rather evil thing to do given the potential for misuse. Hopefully there will not be any phenotypic information provided, but if they somehow manage to run out of money, who knows? Not all phenotypes require a whole organism, as I'm sure you know.
Also, I'm not saying "wait until we know everything there is to know". I'm saying "let's not go off half-cocked -- didn't we learn anything from the schizophrenia 'gene' study?". I am more than a little concerned about this.
Selling the information along is another thing, I'm just responding to your comment I'm not saying 23andme are a great company.
I've noticed myself that when I know a bit about an area, its easy to look at a particular company and see how far behind the research they are. 23andme are trying to take genome sequencing and give it to the masses. Is it very useful? probably not, but they are trying to hit a consumer price point - so doing a whole bunch of original research on genome sequencing, or making a sequencing chip of their own just isn't practical.
A lot of Brin's money is and no doubt will go to good causes, but this isn't suppose to be a charity. Even if it doesn't take off, it's likely to be a positive thing for the personal genome industry not a negative.
I thought 23andme was using an Illumina chip? There was a press release stating that they partnered with Illumina.
Certainly consumer level genetic testing has a long ways to go to become truly medically useful. But 23andme will be well positioned to make the transition to medically useful whole genome sequencing 10 years from now when it is economical.
Ah, you are correct. 23andMe is using Illumina, Navigenics is using Affy. I have my doubts as to whether either approach is truly useful at this point in time to non-researchers, but I could be quite wrong. My objection to 23andMe and Navigenics and so forth is based on the inherently incomplete information provided by a static snapshot of a person's polymorphisms.
As far as email -- sure. Though I don't know whether I'll be pleased to be associated with such a screed when the full details of their plans are more apparent.
Selling their database to pharma companies is exactly the sort of evil I'd hoped they'd avoid. Oh well. Looks like the Google way of thinking didn't infect 23andMe after all.
I can't find the link on their site now, maybe it's gone, but I also believe that 23andme is using a modified Illumina chip that is more inaccurate than the standard chip but is 10 times more likely to return SOME result.
Not sure that it matters -- if you have a less accurate chip, you just resequence (run it) to greater depth. (If you can afford to, that is... and in all likelihood, they can) That's effectively what Helicos et al. are doing anyways. Their call rate for whole-genome resequencing is on the order of 99.7% at a depth of 2; at a depth of 20-50 maybe it becomes useful.
If you think of the chip (hybridization) methods being depth of 1, eg. a single scan per sample across however many probes there are for a given SNP, then you can probably see where the synthesis methods win. It's really cheap to re-run them. It's twice as expensive to run 2 chips...
All of the sequencing-by-synthesis methods are less accurate but cheaper and ultimately more useful than anything based on the existing hybridization arrays (or beadarrays). If you look at eg. the Personal Genome Project, there is a good summary from the Church lab about how eg. padlock sequencing is different from Affy/Illumina/Nimblegene hybridization arrays.
Also note that (1 - 0.999) * 550K still gives 550 miscalls.
D'oh. (Another big win for sequencing by synthesis -- just do it again, and again, and again...)
This is what I'm getting at when I bitch about 23andMe not advancing the state of the art. With the money they have, they could. But they're not.
People generally shell out a lot more than that for many tests.
I thought it was pretty cool. The only reason I'm not paying is that they do not seem to have a service to keep my info updated based on new discoveries.
Yeah, but there's a reason that traditional genetic tests are expensive: they're carefully developed, calibrated and tested to tell you something reliable about a specific gene or condition (they're also usually FDA-regulated, and therefore subject to clinical trials. This is not a trivial expense. It also happens to be an expense that 23andMe has bypassed.)
If a regular genetic test is equivalent to obtaining the contents of a single page in a very large book, what 23andMe is doing is closer to quickly skimming the index. It's cheaper to do on a large scale, but much less detailed. It's therefore a vanity expense, which has to affect the price point.
This is all well and good, I suppose. You have a service of fairly dubious utility being offered, to a bunch of people who likely overestimate its utility. In other words, it's just like any other service in that respect. Maybe less useful than most. (edited to remove erroneous Affy/Illumina references -- with that sort of an error, you better not regard any of the rest of this post as definitive either!)
What sucks, in my mind, about this situation is that the company doesn't seem to be driving any real innovation. Right now, you can have your genome run against a dense panel of SNP markers, and find out what it contains at swab time. That's non-exhaustive, but statistically it barely matters because if you have major new mutations in conserved genes, you're probably dead. Most such mutations are fatal. The ones that aren't continue to be passed around as time goes on. But there are a number of other influences on the way your body ages, gets diseases (or doesn't), passes diseases on (or doesn't), and so on, which we DO know about. First off is epigenetic influences such as whether some of the cytosine bases are decorated with methyl groups. If 23andMe has any intention of offering new scans as the methylation chips become more useful, they're not telling. It will be some time before the success of genome-wide association scans, such as the successful efforts to locate SNPs and SNP (haplotype) blocks which put a person and their offspring at high risk for Crohn's colitis, is wide-spread enough that the genetic information is useful to the average person. The epigenetic and environmental information carried in a person's cells, and the interaction effects of all 3? Could be decades in the most optimistic case. If you think brute-force computational power will solve that, let me know how your protein folding programs are doing. Brute force CPU power was supposed to solve that, too.
Meanwhile, companies like Helicos and Pacific Biosystems are producing technology to cheaply and completely resequence a genome using whole-genome amplification from small samples. They're getting smacked around by the bear market in biotech, but if they succeed, and it looks like they will, then instead of a panel of 1000K SNPs, you'd get ALL of the information in a snapshot of your genome. All 3 billion bases, maternal and paternally derived copies -- remember, you can have different maternal and paternal alleles -- which means all the repeats, microsatellites, and other bits that the SNP chips don't get. (There are plenty of other gotchas as well -- but this is getting really long and I'm not in a mood to edit)
About the use of the word ''snapshot'' -- it's because cellular DNA appears to change in response to cellular RNA manipulation. Tom Cech won the Nobel prize for showing that RNA by itself could store, manipulate, and transmit information at a cellular level. More and more, it's appearing that RNA is actually running the show even in the presence of DNA, and DNA is more like short-term storage for the transient instructions from the RNA. So the DNA, rather than being static, is not just getting methylated and wrapped into bundles, but literally changed on the fly. (If the preceding story is correct -- which no one seems to know)
So what they're assembling at 23andMe is a big tissue bank, a collection of mailed-in buccal swabs, and they're offering the (IMHO, dubious) service of guiding you through that information. There are some useful things you can learn from your genome even now. If you and your spouse have certain mutations -- say, you're both carriers for Tay-Sachs -- then you know right off the bat that 25% of your children are likely to have the disease. If you have certain variations in, for example, TCF7L2 (see http://www.nature.com/nature/journal/v445/n7130/full/nature0...) your risk of type II diabetes is increased (though this is not nearly as conclusive as studies such as those for Crohn's). But most diseases aren't as simple genetically as Tay-Sachs or cystic fibrosis. Most are complex, apparently the product of environmental, genetic, and ''other'' influences. So what you are NOT going to find out definitively is what treatments you might respond to, WHETHER you are more likely to die of colon cancer than anything else... not solely from 23andMe. Not at this point, when there is a very great deal of uncertainty even within the statistical genetics community as to the overall utility of many of these expensive genome-wide association studies. Are they producing results? Fuck yeah! Are those results directly translating into cures or treatments? Uh...
If the current perception is correct, and RNA + environmental/epigenetic influences are in fact changing the genome on-the-fly, then a tissue bank that can re-run the snapshot you sent in (as a buccal, aka cheek, swap) against newer technologies may be of limited utility. The inherent utility is, in that scenario, delivered by the newer technology, and the expertise to run a 'hot' scan on affected, unaffected, or potentially-affected individuals at a particular point in time.
It's unclear to me how their infrastructure could help with this final item, and for all of Brin's wealth, I still don't see them bringing humanity closer to that point. The information could hurt you, by making you uninsurable, but I can't see a snapshot of potential risks helping people the way that newer technologies offering time-lapse snapshots of a person's genome may. That's why it bugs me.
Had to get that off my chest, I guess.
JMHO.