Should you trust a genome scan?

(Newsweek reporter Mary Carmichael has a DNA dilemma: should she buy a direct-to-consumer genetic test? To help answer that question, she’s recruited people with expertise in various areas related to personal genomics – and a diverse range of opinions about the industry – to address specific areas of concern. At the end of the week she’ll announce her decision.

This post is my response to one of Mary’s questions: How does she know if she can trust the results, and should she be scared of what she might find out? A summary is also posted on the Newsweek website along with answers from Thomas Goetz, Hank Greely, Robert Green and Misha Angrist.

Genomes Unzipped is well-represented in Mary’s project: Jeff Barrett wrote about risk predictions yesterday, and tomorrow our resident legal expert Dan Vorhaus will be discussing the present and future state of regulation of the DTC genetic industry.)

The results you receive from genetic testing companies rely on two critical steps: firstly, the generation of your raw genetic data; and secondly, the interpretation of that data into information about your ancestry, family and disease risk.

For reputable genetic testing companies – and I would count the four major personal genomics companies (23andMe, deCODEme, Navigenics and Pathway Genomics) in this category – the first step is generally extremely accurate. These companies rely on the same technology used by academic researchers studying the genetic basis of human disease, applied in carefully quality-controlled labs, so their error rate is typically very low. As an illustration, I recently had an opportunity to compare the raw genetic data provided by two companies to Times journalist Mark Henderson, and found an error rate per company of around one in every 14,000 data points: that’s far better than most routine clinical tests.
For a company like 23andMe with a very active participant community there’s also an extra layer of protection: feedback from other, often highly-informed, customers. When the lab that carries out 23andMe’s genetic testing recently made a mistake that resulted in incorrect data being uploaded for more than 80 customers, there was an immediate flurry of comments about strange results on the company’s discussion forum that resulted in the problem being identified and the affected results being withdrawn within 24 hours. As Dan Vorhaus argued at the time, while the mistake could have been made by any genetic testing lab (including a clinical one), the rapid response to the error was to some extent attributable to the direct-to-consumer approach of 23andMe:

It is worth considering, for instance, whether the same sample swap would have been as quickly identified and addressed – particularly in the case of a “passive recipient” – had the genetic results been delivered to 96 separate doctor-patient pairs (with the patients possibly unable to access the underlying genetic data at all). At least in this particular example, the inevitability of genetic testing errors argues in favor of more consumer access – not less – to allow individuals to continue to play an active role in finding and correcting such mistakes.

The second step (interpretation) is much trickier. Analysis of large-scale genetic data is still a new field, and the most accurate ways to convert genetic data into useful information are currently unclear. While reputable personal genomics companies currently do better job of interpretation and visualisation of these complex data than anyone else on the planet, their analyses are still far from perfect.

Still, imperfection and uncertainty is the price you pay for gaining early access to the cutting edge of any new technology. If you’re happy to wait two or three years to get access to your genome you will get a far more polished product: you’ll be able to get access to a far larger proportion of your DNA than the small fraction currently analysed by personal genomics companies, and the genetic and environmental contributors to complex disease will have been more convincingly untangled.

In the meantime, here are three rules of thumb that are worth keeping in mind before embarking on any personal genomic test.

Firstly, choose your testing company wisely. Here, apply the same approach you would for any major purchasing decision: Google the name of the company and look for positive and negative reviews; as an obvious filter, look at the company’s website for the signs of dodginess you’ve come to expect from any fly-by-night online scam (poor grammar, dubious customer testimonials, offers to sell you nutritional supplements based on your genetic data). A legitimate company should provide you with a demo version illustrating the type of information you would receive as a customer, giving enough context for you to understand the results, as well as extensive links to further information. Ideally, it should also provide an easy way to download your raw genotype data, allowing you to perform your own analysis on it if you so desire.

Secondly, and perhaps most importantly, you should engage with your data: dig deep, read as much as you can, and ask questions about anything that doesn’t make sense to you. Right now the data provided by personal genomics companies isn’t profoundly useful from a health perspective for most of us, so should treat this process first and foremost as a learning experience. You’ll learn the most if you approach your own genome with a skeptical eye, searching for inconsistencies and surprising results and then tracking down their origins. This process may also alert you to errors in your data or its interpretation, as journalists Peter Aldhous and Mark Henderson (subscription required) have illustrated. Read beyond the data provided to you by the company itself; to help build up background knowledge, you could always follow us here at Genomes Unzipped.

Finally, if you find something you think is truly worrying, follow up with an independent expert (a doctor or genetic counsellor for health-related matters) to have the results confirmed and put in context before making any life-changing decisions. These aren’t clinical tests (and aren’t labelled as such); you should treat them as providing you with information, not definitive answers, and certainly not diagnoses. However, be aware that most doctors are ill-prepared to deal with the types of complex information provided by a genome scan, so make the process as easy for them as possible (most companies provide a simplified print-out of results you can present to a clinician).

Should you be scared about receiving your results? This is a tough question to answer, since the range of information offered by DTC genomics companies is so diverse, and different people respond to their findings in different ways. For what it’s worth, studies so far suggest that genetic disease risk predictions, even for serious conditions, cause surprisingly little long-term emotional distress for recipients. However, other results – relating to paternity, for instance – could have a profound and irreversible impact on your life.

My personal view: as a scientist, my desire for information – however incomplete and imperfect – outweighs my fear of the unknown. However, not everyone feels the same way; this is a question every individual simply needs to answer for themselves.

  • Digg
  • StumbleUpon
  • del.icio.us
  • Facebook
  • Twitter
  • Google Bookmarks
  • FriendFeed
  • Reddit

4 Responses to “Should you trust a genome scan?”


  • A nice overview! Just one response:

    > – – if you’re after perfect confidence, you could always wait two or three decades until the genetic and environmental contributors to complex disease are more completely untangled.

    Among the possible time scales and benefits of waiting, I’d suggest that this presents an arbitrary and too distant target. No one interested enough to read this is going to want to put their interest on the shelf that long. However, the next two or three *years* will likely show huge changes, and it might be worth waiting that long. What will doubtless change in that time is both the scientific substance of complex disease risk prediction, and the extent to which such estimates are in fact broadly marketable. Either could change in either direction, but I’m guessing 2-3 years is a good guess for when much of the dust will have settled. If this sounds too cautionary, note that the last several excellent summaries posted here have been about limits to the present state-of-the-art. In the mean time, those concerned about specific familial traits can always get existing, targeted tests.

  • Daniel MacArthur

    Hi Stephen,

    Yes, fair point – that paragraph doesn’t make my point clearly at all. I’ve rewritten it to be more clear; and I agree that your 2-3 year time-frame is a reasonable one to select.

  • G-Unit-Test

    > However, the next two or three *years* will likely show huge changes

    But it’s not at all clear that the interpretable portion of the genome will rise at anywhere near the speed that we can pull in raw bases. The vast majority of that sequence will be a research project for the foreseeable future.

  • A good balanced statement.
    On the matter of time its a bit like the question on how long is a piece of string. Its all in the perspective of the one asking the question.
    If you have the curiosity follow you inquiry but make your own investigations on what you can expect.
    I think I research for more than 12 months before I took the first DNA test. That now is many tests ago. As I learned and understood so my confidence ingreased. I was always aware that this was an evolving science and that was was true today may change tomorrow.
    Initially my first test were all YDNA or mtDNA, but the last was a multi chromosome test.

Comments are currently closed.

Page optimized by WP Minify WordPress Plugin