A huge price cut by personal-genomics pioneer 23andMe may well herald the demise of the commercial personal-genomics business before it even really got started -- and for much the same reason that startups such as Celera Genomics saw their genome-based information businesses crash and burn almost a decade ago.
Earlier today, 23andMe slashed the price of a rough-and-ready genome analysis by 60 percent, to $399 from its intial price of $999. The Google-backed company says it's "democratizing" the personal-genomics business, where prices for similar genome scans have ranged from $1,000 to $2,500, thanks to a new high-density Illumina gene chip that makes it possible to identify more genetic markers at a lower price.
At first glance, this looks like a straightforward example of Moore's Law translated to genetics, and that's exactly how many in the blogosphere have enthusiastically described it. Wired.com's Aaron Rowe, for instance, argues that the 23andMe move makes personal genomics "a viable hobby" well within the reach of "cash-strapped grad students, frugal genealogy buffs and other not-so-early adopters."
If so, that's all to the good, and you'd expect major 23andMe competitors such as deCODE Genetics' deCODEme service or the "other" personal-genomics startup in Silicon Valley, Navigenics, to swiftly follow suit. (For a description of what exactly these services do, see the end of this post.) And maybe they will, although probably not for the reasons you might expect.
The main point to remember is that personal-genomics companies don't intend to make money by selling the tests. Instead, their business generally depends on amassing a giant anonymized database of customer genetic information that can be mined for research studies by academic researchers or drug companies.
With this in mind, you could look at 23andMe's price cut as a sort of loss leader -- an attempt to jump-start the data collection in order to kick the real money engine into gear. And that would make perfect sense, but for two basic problems.
First, the research model of personal genomics has always been a dicey proposition. To be useful for research into genetic influence on disease, for instance, genomic information has to be paired with so-called phenotypic data -- basically, a rigorous description of an individual's health, family background and important behavioral and environmental factors (smoking status, prescription and recreational drug histories, and so forth).
But it's never been clear exactly how 23andMe planned to extract phenotypic information from its customers. It's even harder to see how this data would prove useful in research, since any disclosures would be entirely self-reported by a self-selected fraction of all 23andMe customers -- two classic sources of research bias.
Second, falling prices for SNP scans -- and, eventually, for much more informative full-genome readouts -- means that academic researchers can now afford to conduct large-scale genomics studies on their own. In fact, this is exactly what the nonprofit Coriell Institute aims to do with its Personalized Medicine Collaborative, an ambitious effort to take SNP scans of 10,000 people by the end of next year -- for free. (Ultimately the center plans to scan 100,000 people.)
Now, research institutions like Coriell aren't going to drive personal-genomics companies out of business simply by undercutting the price of SNP scans the way ThinkGene's Andrew Yates suggests:
The good news for genomic testing DTC competitors 23andMe, Navigenics, and deCODEme is that they were the trailblazers and that Coriell is following their lead. The bad news is that another word for "trailblazer" is "cannonfodder." Sure, you might get a statue, but you're going to be dead. (It's typically bad for business when an established, well-funded, experienced competitor releases a better version of your product for free.)After all, some people will still want immediate access to their own genetic information, and would be willing to pay for the privilege. No, the real threat is that cheap SNP scanning will undermine the very research studies that companies like 23andMe and Navigenics planned to make their bread and butter. (To his credit, Andrew almost argued this point in a subsequent post, although he still seemed to think the startups plan to make money by selling their tests, instead of access to their data.)
Put it this way: Once reputable medical institutes get into the genomics-research game in a big way, research scientists aren't going to pay out significant fractions of their grants to get access to a commercial database when noncommercial databases -- quite likely of higher quality -- are more readily accessible. Drug companies might be more willing to pony up in the short term, but only until they can access better and publicly available genetic-variation databases.
This is exactly the dilemma that Celera, which raced the government-funded Human Genome Project to a draw in sequencing the first human genome, faced earlier this decade. The startup had figured that it would provide a better annotated and generally more useful gene database that companies and research labs would gladly pay to access. The opposite proved true, as the contributions of thousands of individual researchers turned the public databanks into one of the first successes of open-source biology. Celera soon abandoned the data business, and after a failed attempt to turn itself into a drug company eventually morphed into the genetic-diagnostics business it is today.
Something very similar seems destined to happen to personal-genomics companies that can't radically retool their business models. Startups like 23andMe are dependent on regular infusions of cash provided specifically in hopes of a big payoff down the road. If hopes of that payoff disappear, so does the venture capital.
Technical addendum: All of these personal-genomics services scan roughly 500,000 to one million DNA base pairs -- technically, single-nucleotide polymorphisms, or SNPs -- from a genome to provide a tentative outline of what exactly makes an individual genetically distinct. Customers provide DNA samples by mailing a spit-filled tube into the company, after which they can poke around their SNP map over the Internet to check out their likelihood of developing various diseases or their degree of relatedness to other people, if they so choose.
These SNP maps fall well short of a full readout of the six billion base pairs that make up the full human genome. There's a raging argument as to whether the SNP readout is useful enough to warrant the expense, with genome enthusiasts insisting that any information is better than none while some doctors and state regulators argue that most consumers have no way to make practical use of the data.
Image of the printed-out human genome via Flickr user JohnJobby, CC 2.0