The importance of confusion

Steven Pinker's recent article in the NY Times is a rich source of insight into the field of personal genomics and the experience of personal genomics customers - if you haven't read it already, you really should.
This paragraph, for instance, seems to perfectly encapsulate the experience of the average intellectually curious personal genomics customer:

It became all the more confusing when I browsed for genes beyond those on the summary page. Both the P.G.P. and the genome browser turned up studies that linked various of my genes to an elevated risk of prostate cancer, deflating my initial relief at the lowered risk. Assessing risks from genomic data is not like using a pregnancy-test kit with its bright blue line. It's more like writing a term paper on a topic with a huge and chaotic research literature. You are whipsawed by contradictory studies with different sample sizes, ages, sexes, ethnicities, selection criteria and levels of statistical significance. Geneticists working for 23andMe sift through the journals and make their best judgments of which associations are solid. But these judgments are necessarily subjective, and they can quickly become obsolete now that cheap genotyping techniques have opened the floodgates to new studies. [my emphasis]

I specify "intellectually curious" because there are probably some personal genomics customers out there who simply accept the risk predictions they get from 23andMe or deCODEme without much further introspection, but I'd guess such users are in a substantial minority. Given the type of person who is likely to fork out money for a genome scan, I suspect most customers end up doing some digging into their data - and when they do, they are inevitably faced with the same tangle of paradoxes and contradictions as Pinker was.
This confusion arises not because of poor judgement on the part of 23andMe's geneticists, but rather because our current understanding of the genetic basis of most commonly variable human traits and complex diseases is still incredibly primitive. Despite the hundreds of associations between common variants and complex diseases like type 2 diabetes identified over the last two years, the risk predictions available for most (but not all) of these diseases are currently barely better than noise. That means that current risk predictions can vary considerably between companies, and from the same company over time, due to changes in prediction algorithms and the addition of newly-discovered risk variants.
This situation is temporary - over the next decade, as our knowledge of the molecular basis of complex diseases grows, personal genomics customers will see their risk predictions continue to fluctuate before slowly converging on some true level of genetic risk. If personal genomics can be successfully combined with high-resolution data on environmental risk factors, we will then be looking at predictions with substantial clinical and preventive utility.
OK, so personal genomics will be fantastic in ten years' time - what about now? Does the meagre predictive value of current genomic disease risk predictions mean that people buying a genome scan right now are wasting their money? I'd argue not, for several reasons.
Firstly, there are several complex traits and diseases for which current gene-based predictions *are *useful: for instance, the APOE variant associated with late onset Alzheimer's disease (which Pinker deliberately excluded from his own analysis, saying he didn't need a boost to his "current burden of existential dread"), or the genes associated with age-related macular degeneration. If it's ominous news you're after, you can still luck out with current genome scans.
In addition, the variants assayed by personal genomics companies provide surprisingly powerful information about genetic ancestry, a topic of considerable interest to an extremely large audience (although it must be said that personal genomics companies are yet to exploit this information to anywhere near its full potential). While we wait for the genetic underpinnings of most complex diseases to be unravelled, ancestry can even provide rough surrogate information about genetic risk due to variation in disease risk between human populations - a particularly useful service to those who currently know little about their own deep origins (such as some adoptees).
But perhaps the most valuable contribution of personal genomic data stems, perversely, from the confusion itself: the sheer complexity and uncertainty of the data encourages customers to explore the complex interface between genetics and health for themselves. Few experiences could strike as heavy a blow against someone's intuitive notion of genetic determinism as looking through a list of probabilities in 23andMe's risk report. In the quote above, Pinker memorably describes the experience as akin to "writing a term paper on a topic with a huge and chaotic research literature" - a frustrating experience that few people would volunteer to perform unless, as in this case, the subject was themselves.
To use an analogy: personal genomics provides a distorted and fragmented mirror, and in the process of peering around to catch a glimpse of our own faces we instinctively learn about the flawed topology of the mirror even as we learn about ourselves. In more prosaic terms, every customer of a reputable personal genomics company is one less person who will fall for the next simplistic "gene for erectile dysfunction" story peddled by their local newspaper, or believe the absurd claims from DNA Dynasty and its ilk.
My point is that whatever the long-term health impact of the genomic revolution, the creation of a community of well-informed, genetically literate individuals will be a lasting legacy of the personal genomics industry. Hell, even if it turns out that the genetic architecture of most human traits is just too complex to ever make accurate health predictions, at least we'll have a broader community of people who understand that this is the case.
Pinker's own assessment of the future of the personal genomics industry is lalso optimistic:

Personal genomics is here to stay. The science will improve as efforts like the Personal Genome Project amass huge samples, the price of sequencing sinks and biologists come to a better understanding of what genes do and why they vary. People who have grown up with the democratization of information will not tolerate paternalistic regulations that keep them from their own genomes, and early adopters will explore how this new information can best be used to manage our health. There are risks of misunderstandings, but there are also risks in much of the flimflam we tolerate in alternative medicine, and in the hunches and folklore that many doctors prefer to evidence-based medicine. And besides, personal genomics is just too much fun. [emphasis in original]

The key, of course, will be ensuring that personal genomics companies continue to provide information that is accurate and sensibly portrayed. Regulators will have a role to play there, although it's likely to be a clumsy and heavy-handed one. More importantly, I suspect, will be the policing provided by the community of personal genomics customers: by sharing information, highlighting contradictions and generally questioning everything they hear, they are likely to be a far more effective regulator than government agencies could ever be.
Such policing won't do away with confusion - confusion is simply a fundamental feature of the world of human genetics circa 2009 - but at least it will ensure that this confusion is conveyed accurately to consumers, in all of its deep, messy, incomplete, paradoxical and tantalising glory.
Subscribe to Genetic Future.