Daniel MacArthur has been indispensable lately -- if you are interested in consumer genetics, check out his post on Amway's "Heart Health" genetic tests. The post refers to an article in Technology Review, and ends:
These are the types of tests that give direct-to-consumer genetic testing a bad name; careful regulation that punishes these bottom-feeders while leaving scientifically supported tests intact (i.e. the major personal genomics companies) is long overdue.
The Economist has a "special report" on personalized medicine, focusing on the business of the current set of sequence providers. Generally speaking, Dan MacArthur (Genetic Future) is quicker on this beat than me, and today is no exception, as he posts on the article:
The ruthless competition in sequencing technology mean that companies in the sequencing industry will spend the next few years butting up against the bottom line, struggling to cut the price of a genome by a few more dollars to match their competitors. That's a tough way to make a living. On the other hand, companies with skills in interpreting genomes - like, say, 23andMe and deCODEme - will be in a good position to take advantage of plummeting sequencing costs to provide useful genetic information to consumers.
There are some other business models floating around out there, but MacArthur is right about the dichotomy emerging. Rapid high-throughput sequencing is a big business with little direct marketing potential. You want to keep your machines running full tilt -- it's one of the challenges at even the scale of Research I universities, where these resources are purchased as a collective and shared out across many departments. So if you're in the business of providing sequence, you want to contract out in large chunks, not do little teeny bits of preparation at a time. There may well come a time when the volume is so high that a provider handles both ends of the process -- Amazon-like -- but that will be awhile.
Meanwhile, nobody's quite figured out how to sell sequence to people. Some of these companies have had some great ideas -- I think 23andMe's effort to tap into the pregnancy/early childhood community is very shrewd. But the reactions of "experts" quoted in the Economist article are pretty telling:
Though he has the world’s most advanced gene-sequencing technology at his fingertips, Dr Altshuler refuses to get his own genome sequenced: “If someone gave it to me on a CD, I’d refuse to look at the disc. The information is meaningless.” Bill Gates agrees. He has not had his genome sequenced either, nor does he plan to, though after a moment’s reflection he adds, “unless I find out I have cancer.”
Yummy. Cancer. I smell some great marketing there...
Let's face it, there's only so far you can blackmail people into buying your service. The Mafia doesn't inspire great brand loyalty. The overtures to the pregnancy community are very smart, but that's still marketing based on anxiety. Turning anxiety into a positive is a great idea
I used to feel the way Altshuler does. What good could it possibly do to have my genes sequenced? I know the limits on the usefulness of the data. There's minimal medical value for most people right now. I'm not even sure what I would do with them from a genealogical/historical perspective. If I learn something from them, it's likely to be either bad (some T2D risk allele) or something I already know (HERC2 genotype). What a buzzkill.
I can see why some experts are skeptical of the value of widespread genetic testing. People are being taken in by Brinkley-esque schemes, companies are out there right now taking money for "genetic tests" that aren't actually testing any genes. The field of "nutrigenomics" has spawned several such schemes, according to a 2008 article in Scientific American. Genetic tests are not likely to give anybody actionable information about their health, unless they are looking for very specific genotypes with high penetrance, like BRCA1 cancer-associated mutations. Maybe with a lot more knowledge about multilocus risks, we'll get to the point where genetic tests are as informative as cholesterol tests -- far from perfect, but at least useful indicators of risks. But we're not there yet, far from it.
Still, I have to say my attitude has changed in the last few months. I don't know everything I might do with the data, but I'm starting to get some good ideas. I've been talking to a lot of audiences, and I have to say people are really interested and excited to hear stories about genes.
There's an audience there, willing to hear about genes because of what it tells them about themselves, about their history. "Interpretation" isn't the service -- content is the service. Right now, they seem to be treating the business model basically the same as those people who warn you when somebody changes your credit rating. Just wait around, and when we learn something new about your genes, we'll give you an update.
I think the right model is the premium satellite channel model. What does it take to be the HBO of genomics? It takes compelling content -- new, fascinating content that people are willing to subscribe to. There ought to be a lot more video. It ought to connect people to subjects they already know are deep and important, like (if I could be so bold to suggest) our evolutionary history. The content is the hook -- it gets people thinking about how interesting the data are, how they connect them to other people, how much they'd like to know more.
A multi-hundred dollar purchase is a very high barrier to entry. Where's the $25 service that someone can buy to get involved at a lower level? Where's GENOME the magazine? These are loss-leaders for the premium services. How much would it cost to get people to write compelling stories about the histories of common haplotypes? Every month you have a genealogy feature story, a health feature, an evolution feature, maybe comparative genomics. The magazine's features and the videos all have hooks into the software application, which tells you which of your genotypes are in the stories.
The point is, you can't just give people a database. You have to curate the information, select it and then present it. You don't need a talking paper clip -- you need a Robert Osborne. Or several hosts -- an Alton Brown, a Tony Bourdain, heck, even a Bear Grylls. Except people who really know their genetics and know how to explain stuff clearly. A slate of basic-cable programs on genetics, maybe in collaboration with an existing channel, each pointing out the ways that the company's customers have underwritten and participated in the research. Start with the fun stuff -- not the high-falutin' "African-American Lives"-type presentation we usually see. Each show should have parts that draw people to the website -- maybe haplotype-specific stories, maybe real footage of customers encountering their personal histories.
This aspect of presenting and communicating science is the true challenge today for many scientists and journalists. It's not easy to do this stuff well. But I think in the area of genetic testing, it's not an optional add-on -- any valid business model is going to depend on keeping people engaged as research continues. Good stories give you a chance to sell your product. This week, more than 20 million people downloaded the video for one woman's audition on "Britain's Got Talent." If she had a CD ready, how many would she have sold?
I was reading this story about "genetic surveillance" by law enforcement. I'll blog about it later.
In the meantime, I had this idle thought. Suppose you got a bunch of genotypes from some testing company and started writing about them on your blog. How many would it take for law enforcement to be able to identify you?
It happens on CSI all the time -- they find some rare chemical or pathogen or gene in a sample from a crime scene, and that's the crucial clue that leads them to the killer. So, if you happen to be a chimera, if your body has absorbed your fraternal twin, or if you've had a bone marrow transplant, you should expect that Grissom is going to catch you.
The principle of DNA fingerprinting is that you put together a bunch of relatively common alleles, and the combination of them is so rare that it identifies you uniquely. You don't have to depend on a vanishingly rare allele, you just have to depend on Mendel's Law of Independent Assortment. CODIS uses 13 STR markers, for which the alleles are relatively rare, giving a probability of one in billions that two people would share the same markers.
But if you just wanted something probative -- maybe the investigators have a DNA sample, and they are just checking your blog to see if they should look at you further, you don't need anywhere near 13 STRs. If you're blabbing about your genotypes, the frequency of a common genotype in the population is going to be anywhere between 10 and 50 percent. Let's say the average is 30 percent, like the frequency of blood type A, or lactase heterozygotes in the US. Well, five genotypes like that will give a chance less than 1 in 400 that you're the same as the suspect by chance.
Good enough for a warrant? If you're a type AB positive, lactase heterozygote, blue eyed redhead who blogs about her APOE status, that's seven, and some are pretty rare. Definitely less than one in 4000. That would be good enough for CSI!
Unless it turns out you're secretly a man.
A reader wrote me today:
I am wondering if you could suggest a general article for me to give to people who tell me how exited they are about sending a cheek swab to National Geographic for analysis. I imagine you have written something like this, but don't remember where it is. Any thoughts?
I wrote back, and thought I should share. I have an article in Slate from a couple of years ago that covers that topic:
http://www.slate.com/id/2138059/
Also, there was a NY Times article in late 2007 that basically covers the same difficulties:
http://www.nytimes.com/2007/11/25/business/25dna.html
I blogged about that article at the time. I also have a post from the time of the Slate article.
And my "testing" tag leads to a bunch of posts on the topic of genetic testing. I've never really sat down to write an FAQ on the topic, but I think it would be worth doing. Genetic Future does a very good job of covering current commercial offerings in genetic testing, a beat also covered at Eye on DNA.
If someone was thinking of sending $100 to Genographic, I have nothing against it as a form of entertainment. I think the information you get that way is, for most people, very superficial and unsurprising. Basically, they're giving you your paternal Y chromosome lineage if you're a man, and your mtDNA if you're a woman. That fee pays for your test, the information they give out to you as a participant, and an extra amount that they will use to subsidize testing of various populations around the world.
But if you're looking for the entertainment value, you will be much better off saving up $400 to send to 23andMe. That pays for a whole-genome SNP survey, including various phenotype predictions. They have all the genealogical information from the Y and mtDNA that you would get from Genographic, and it's better-presented. Overall, it's a much better deal, and more likely to generate something interesting to share with your friends. Of course, it costs more, but it's a better value.
Now, for serious medical purposes, or for trying to do phenotype predictions of your children, I don't recommend 23andMe. I honestly think that gene testing now is best treated as expensive entertainment. You might get a laugh, or learn something about your ancestry that you didn't know. But you won't likely learn anything that would be worthwhile to your children. Not yet, anyway. This is not an investment.
In other words, don't spend your food money. Think of it as eight months of cable TV.
Honestly, that was my first reaction to an article that includes relatively neutral grey-background shots of Pinker from several angles. Way to go, dude!
The article is Pinker's insider's account of the Personal Genome Project, and is larded with some ongoing results in human behavioral genetics. Both should interest those who have been following technology and human genetics. Earlier this week, I posted about Sharon Begley's reaction to Pinker's last Edge essay; I thought the short section cited probably didn't reflect the full nuance of Pinker's views (for instance, as expressed in The Blank Slate). Today's essay in the NY Times Magazine does a better job of describing the science, along with its possible benefits and risks:
Though the 20th century saw horrific genocides inspired by Nazi pseudoscience about genetics and race, it also saw horrific genocides inspired by Marxist pseudoscience about the malleability of human nature. The real threat to humanity comes from totalizing ideologies and the denial of human rights, rather than a curiosity about nature and nurture. Today it is the humane democracies of Scandinavia that are hotbeds of research in behavioral genetics, and two of the groups who were historically most victimized by racial pseudoscience — Jews and African-Americans — are among the most avid consumers of information about their genes.
Pinker describes the current state of what behavior geneticists know (most things are heritable, shared familial environment accounts for little variation) and what they don't know (which genes account for any of the heritable variation). Given this state of knowledge, the results from direct-to-consumer genetic testing seem to approach the trivial -- and are notable for their exceptions more than their rules:
Direct-to-consumer companies are sometimes accused of peddling “recreational genetics,” and there’s no denying the horoscopelike fascination of learning about genes that predict your traits. Who wouldn’t be flattered to learn that he has two genes associated with higher I.Q. and one linked to a taste for novelty? It is also strangely validating to learn that I have genes for traits that I already know I have, like light skin and blue eyes. Then there are the genes for traits that seem plausible enough but make the wrong prediction about how I live my life, like my genes for tasting the bitterness in broccoli, beer and brussels sprouts (I consume them all), for lactose-intolerance (I seem to tolerate ice cream just fine) and for fast-twitch muscle fibers (I prefer hiking and cycling to basketball and squash). I also have genes that are nothing to brag about (like average memory performance and lower efficiency at learning from errors), ones whose meanings are a bit baffling (like a gene that gives me “typical odds” for having red hair, which I don’t have), and ones whose predictions are flat-out wrong (like a high risk of baldness).
The second half of the essay focuses on Pinker's own experiences with gene testing and the constraints of behavior genetics. In particular, he discusses the observation that so far the genes found to be significantly correlated with behavioral traits like IQ explain only a very tiny fraction of the heritable variation in large populations -- he calls this "Geno's Paradox". We've plumbed the depths -- as noted here on other occasions -- and if there were any genes explaining large fractions of the variation, we would have found them by now. The observation is easily explained -- the heritable variation is explained by rare alleles or small effects across hundreds or thousands of genes. But this solution means that genome-wide tests will not be good predictors of such traits in the foreseeable future.
With that result prominently in mind, what is the point of all this genome sequencing? Pinker makes the point that the Personal Genome Project is not about predicting phenotypes, it's about research. The participants want to help find new pathways by which genes affect phenotypes. Recording the whole genomes of a well-studied set of people, whose phenotypes have been recorded in more-or-less excruciating detail, is the way to get data for this process. Conceivably, if influential alleles really are rare in the population, we will continue to get valuable data as we expand the set of such public genomes into the hundreds of thousands.
There are many good analogies in the essay. I especially like Pinker's distinction between an person's physical state and the mental state of other individuals who may know genetic information about that person. The conclusion of the essay conveys an important point: Even if the genome were destiny, it's pretty unlikely that any particular gene would explain it:
It’s our essentialist mind-set that makes the cheek swab feel as if it is somehow a deeper, truer, more authentic test of the child’s ability. It’s not that the mind-set is utterly misguided. Our genomes truly are a fundamental part of us. They are what make us human, including the distinctively human ability to learn and create culture. They account for at least half of what makes us different from our neighbors. And though we can change both inherited and acquired traits, changing the inherited ones is usually harder. It is a question of the most perspicuous level of analysis at which to understand a complex phenomenon. You can’t understand the stock market by studying a single trader, or a movie by putting a DVD under a microscope. The fallacy is not in thinking that the entire genome matters, but in thinking that an individual gene will matter, at least in a way that is large and intelligible enough for us to care about.
Well, I've pulled several quotes, but it's a very long essay -- 8000 words. So it's hard to give an impression of the whole thing. I think it will make great reading for the students of my course in genetics. Of course, the printed PDF doesn't include Pinker's legs...
The AP's Howard Fendrich reports on an American Enterprise Institute conference about gene doping:
Gather a roomful of anti-doping experts, administrators, academics and athletes alike — something a conservative think tank did Thursday — and there is no consensus as to whether gene doping, thought by some to be the next frontier in Olympic cheating, is at hand.
Indeed, there isn't even consensus on whether it would be a bad thing.
The article is not really an in-depth coverage of the issue, but merely trades back-and-forth quotes from various conference participants. It's still interesting, though:
John Leonard, executive director of the American Swimming Coaches Association, told of conversations he has had with coaches and scientists in China.
"We are really naive if we are to believe that the Chinese at this point are clean or that they are the only country in the world that is experimenting with genetic enhancement as we speak," said Leonard, who was not a panelist but attended the conference and spoke during question-and-answer periods.
"There are lots of countries in the world who couldn't care less about doing it safely, and there are lots of athletes who will take the chance that they will die in order to win medals. ... Will the United States have the same viewpoint when we start losing gold medals?"
The views of the pro-doping people seem to have been well represented, along with Edwin Moses and others who are anti-doping.
Ronald Bailey opines about coming pressures for politicians to release their genetic test results:
Again, it's just as easy to obtain a DNA sample from a presidential candidate as it would be to get one from a celebrity like Winfrey. Green and Annas are most worried that competing campaigns might engage in "genetic McCarthyism." That is, campaigns will seek to obtain DNA from their adversaries and then release genetic data that suggests that their opponents are somehow unhealthy. Such a tactic could be used to confuse the public because genetic information is easy to misinterpret and to misrepresent. Consequently, Green and Annas argue that "future presidential candidates should resist calls to disclose their own genetic information. We recommend that they also pledge that their campaigns will not attempt to obtain or release genomic information about their opponents." They reject the idea of making it a federal crime to sequence a candidate's DNA without consent. Oddly, Green and Annas overlook the plausible scenario in which some media organization surreptitiously obtains DNA from candidates, and then sequences it and reports the results.
This last scenario, the sneaky approach, pretty much makes it inevitable that genetic test results for future candidates (or anyone else of sufficient interest) will be public. Congress could simply make it illegal to release any record of a person's genetic information. But this would extend the privacy right much further, with respect to publicly obtained information, than it currently goes -- and far into the territory where the First Amendment pushes back.
If it were only politicians, that would be trouble enough. But these kind of data may be used in the same manner as many kinds of "junk science" today. Imagine a custody battle, in which the father hires a private investigator to get a mother's genome. With two variants that yield a 15 percent higher risk of schizophrenia, will the mother's genetic risk be held against her? Or think of corporate boards, looking for a way to dismiss a CEO without paying that golden parachute. Could a genetic test result showing a higher risk for early Alzheimer's give them a reason to invoke a "health" clause in the contract?
Why does this entire topic look like "junk science?" I can think of a couple of reasons. First, genetic information today is essentially meaningless at the individual level. Consider Bailey's description of his own 23andMe results:
The genetic screening company reports that 24 out of 100 people with my genotype will get type 2 diabetes between the ages 20 and 79. The average risk is 21.9 per 100 people. With regard to macular degeneration, 9.5 out of 100 people with my genotype will get it between the ages of 43 and 79. The average risk for people of European ethnicity is 7 out of 100. And 0.94 out 100 people with my genotype will get Crohn's disease between the ages of 20 and 79. The average risk for people of European ethnicity is 0.43 out of 100. I will save for a future article the good news that I also have a number of genetic markers that indicate lower risks for many other conditions. This is the kind of risk information that genetic screening tests will reveal.
These "risks" are based on single genotypes, mostly replicated in more than a single study, but not analyzed with any combination of other genotypes. We simply don't have a way to predict a person's lifetime risk of chronic disease, based on the entire genome. A person who has one known risk allele may have any number of unknown protective alleles. So saying that a person has a XX risk for condition YY is highly misleading. At best, the strongest such variants suggest a higher relative risk, all other things -- including environment, diet, and past disease history -- being held constant.
Second, there is no natural limit to what a genetic test might be claimed to find. Genes affect health, sure, but they also affect personality, IQ, and behavior. Once a person's genetic data are made public, they are hostage to every future study that might show one (or more) genetic variants are associated with some trait.
This is not only about the candidate for President, where the public is judging their future health. This is about a sitting President, whose public genome data show a newfound association with a personality disorder.
Maybe, as George Church hopes, we are headed for a future where these things won't matter. But I think it will be a rough road.
Genetic Future points out the minuses of a commercial gene test offering, which promises to tell parents whether their kids have alpha-actinin-3 gene combinations that are well-suited to Olympic sprinting. Since Dan is an expert on this gene, he can say with some authority that the test isn't adding much value for parents who want to make their sub-8-year-olds into the next Usain Bolt. Plus, for a few dollars more the 23andMe test (and for that matter, other SNP screens) will give you the same information plus many other genes.
Razib, in fewer words, points to the post and says:
Remember those astrology infomercials on TV? "For entertainment purposes only!" Over the next few years many firms will piggy-back on the cultural prestige of science to make a quick buck.
Meanwhile, the Associated Press reports that Obama is likely to "broaden genetics role in medical care":
Obama is also interested in the role that personalized medicine could play as an element of changes in the broader health care system.
"The issue of getting the right treatment to the right person goes with his whole emphasis on health reform," said Mark McClellan, a noted Republican health care expert who served President George W. Bush as Medicare director and head of the Food and Drug Administration. "If we're thinking about reforming the health care system, we should be thinking about what medicine will be like down the road when health care reform is fully implemented," McClellan said.
The article is purely speculative with respect to Obama's actual plans. It does review Obama's efforts as a senator to introduce legislation that would "coordinate the policies of federal agencies whose decisions have an impact on" the use of genetic tests in medical treatment. In other words, will Medicare pay for your genetic test? And should the FDA regulate it? A federal agency capable of getting Medicare and the FDA to use the same protocols for genetic information would seem fairly likely to impose strict record-keeping requirements on commercial gene tests.
Remember that genetic tests presently aren't very useful for predicting complex disease phenotypes, and this problem won't be going away soon. Direct-to-consumer gene testing companies so far seem to be playing the game as if they were selling nutritional supplements: there's some nebulous, undefined benefit related to health, but few specifics. Or as genealogy research aids -- no fear of regulation there. That lies behind Razib's comments about "entertainment purposes only," which I would say is pretty much accurate at this point. There's little or no compelling health-related information in today's genetic tests for ordinary consumers -- which is to say, people who are not otherwise at risk of inherited disorders based on family histories.
From this perspective, Obama has been trying to draw up regulations for something that doesn't exist. But like any good planner, we should consider what is likely to unfold in the future, when we can go beyond the present genome-wide association tests toward interaction analyses of multiple genes and phenotypes. Possibly we'll do better at predicting the risk of complex diseases, and we'll certainly have tests at a fraction of today's prices.
Providing diagnostic value for SNP screens or genome sequences will take a massive effort at standardizing information about joint gene-phenotype associations. Direct-to-consumer gene testing companies presently differentiate themselves based on the different information they provide to their customers. That approach works as long as there is little of value in the results -- the companies today are succeeding or failing on the basis of the communities of customers they are building, with the stories of customers providing the best advertisements. That's the nutrition supplement market.
But that approach will start to fail if genetic tests start to allow serious risk mitigation in health maintenance. If two companies provide divergent information to customers, in a way that impacts the customers' interactions with their physicians, I expect that the outcome will be some massive lawsuits and further federal regulation. If the government becomes the health care purchaser -- and with Medicare it already is the largest -- we can expect to see early federal intervention in this market, focused upon standardizing genetic information provided to physicians.
"Entertainment value" indeed.
Genetic Future has been on fire lately, with various announcements from and about genomics testing companies. More on that later. Today, he reflects upon the publication of two non-European genomes in the current Nature:
So attention has already well and truly turned to converting sequence into biological meaning - and that's a job that will ultimately require many hundreds of thousands of genome sequences, each attached to information about biological traits and disease status. That means the end of the brief era of high-profile "single human genome" papers, which started in a sense with the anonymised, pooled and fragmented human reference sequences published in 2001, peaked with the celebrity genomes of Venter and Watson in 2007/2008, and now ends (I suspect) with two anonymous non-European genomes.
Human genetics now moves into a phase of new challenges and rewards - the era of population genomics.
It's a good post, with the thesis that we will never again see a paper published in a major journal to report the genome sequence of a single individual. I demur for a special case: We'll see fossil and archaeological genome sequences in major journals for quite some time to come. I suppose after a few Egyptian or Neolithic genomes, those also will be reported many individuals at a time. But Pleistocene remains will always be singular.
In the meantime, of course population genomics is what we're all about here in the Hawks lab. Single-locus genetics has gone the way of the dodo. Er...I suppose if you study dodos, you'd better go whole-genome with them, too. My only question: exactly how much hard drive space am I expected to have, if I'm going to deal with 100,000 genomes?
ThinkGene has a nice critical post reviewing some of deCODE Genetics' advertising. The main idea is that genetic tests as yet provide almost no information worth acting upon in a course of treatment.
So, since there is no evidence, only “decision to take action,” you could replace the deCODEme test with a homeopathic test that produces the same ratio of positive results for the same clinical effect. I understand that people sometimes feel they need a special reason to pursue healthcare from which they could always hypothetically benefit, but there is no rational necessity to produce this impetuous prior to treatment other than the psychological comfort of rationalization.
He concludes that some of the claims are essentially the same as magic, since they have no evidence-based support.
