bioinformatics, genomes, biology etc. "I don't mean to sound angry and cynical, but I am, so that's how it comes across"

An improved model for direct to consumer genome sequencing

I read with great interest this story about how a DTC genetics company sent details to a customer about a genetic variant they thought he had, which if true, would have altered his treatment for Crohn’s disease.  As it turns out, the variant could not be confirmed independently and the correct course of treatment was administered.

There will be much hand-wringing about this, and rightly so.  However, for me, it reveals a fault in the model of how DTC genetics and genomics is done.  What happens is that customers provide DNA, and the genetics company provides SNPs and an interpretation.  Of course there are many steps in between those two points, with uncertainty at each stage, and I think separating the “pipeline” is a better model.

So instead of having an all-in-one package, wouldn’t it be better if we had:

  1. Customer buys (and owns) the raw sequence data for their genome (they may end up owning many, from different times in their life or from different tissues).  As it is relatively unprocessed, this would represent the least biased view of the customer’s genome.
  2. The customer can then purchase analysis of their genome separately; including alignment, variant calling and interpretation.  This step would involve education of the customer that different aligners, variant callers and interpretation services have different strengths and weaknesses, and all come with a level of uncertainty.
  3. Of course, as the software and databases improve, step 2 can be repeated on the same data from step 1 over and over again.  If sequencing improves, step 1 can be repeated.

I think owning your own genome data, and buying services to run on that data, is the best model for DTC genetics/genomics.  How would this have helped Dr J?  Well, owning his own data, and having found out that one interpretation of that data suggested a variant which would seriously affect his Crohn’s treatment, Dr J could simply have bought a second opinion on the same data from a different vendor, who would use slightly different algorithms to determine Dr J’s variants.

Thoughts and comments 🙂


  1. OK. So given two (or more) ‘opinions’, which is correct (to the consumer)? One, both or neither…?

    Doesn’t sound any better to me.

  2. You don’t think knowledge that a result carries some uncertainty and that there is an alternative interpretation is better?

    My point is that the way it’s set up, the process and the uncertainty are hidden from the consumer. I’m talking about revealing both.

  3. Consumers don’t like uncertainty. Especially when comes to health/medical matters.

    This is why people still smoke (aside from the addiction aspect). The uncertainty allows them to deny the harm they’re doing to themselves.

    Getting yourself sequenced will be the same. “I feel fine now, so I probably won’t get that cancer despite me having a highly penetrant allele. I still have a 30% chance of being fine.”

  4. Assume that I, as consumer, also want to keep the raw instrument data, so that I am also exposed to the uncertainty of today’s best known methods. I keep the raw data on some central cloud storage, but I keep my sequence with me on my smart phone. Now, imagine, 5minute bioinformatics at the doctor’s office. I come in with an ailment, they pop in my smartphone to look at my DNA, and they provide me with their judgment. If it is serious, we schedule a second opinion, but some process. I like that model a lot, and it would be the same ‘behavior’ as my blood work goes through now.

    And as our BKMs improve, we can revisit the raw instrument data and resolve uncertainty due to instrument, sample, or software. I would sign up for that, as a matter of fact, I’ll do it myself this week: have sample, need lab….

  5. I do not think that decoupling the data pipelines and interpretation would be a solution. A far more powerful proposition is a hosting service that stores your raw data and creates a platform to plugin any vendors of interpretation services… a bit like an app store.
    With apps specific for interpreting specific medical conditions for example… and then the customer pays for those “apps” depending on their needs or on a medical recomendation.

  6. I think that’s just a different way of saying what I said 🙂

  7. I’m not sure I get your point. Consumers may not like uncertainty, but if uncertainty exists, then they should get it – that’s an ethical and moral point.

    As for smoking, there is no uncertainty – it is unhealth, and it will kill you. These are facts.

    People choosing to ignore their data is a different issue to the fact that consumers deserve access to their data, and estimates of uncertainty aroound interpretation

  8. Exactly!

  9. maybe 🙂 … what I meant with my comment is that the customer would have a framework to plugin its data and and manage input and outputs… if you leave it to the customer to find their own best annotation pipeline without a framework … maybe computational biologist would do it … but not the general public.

    The complexity has to be hidden away from the user or it overwhelm him…

  10. Absolutely! I imagined the raw genome sequence data being in some kind of secure “bank”, and then consumers would grant access to their data to companies selling analysis and interpretation under a strict license 🙂

  11. Feeling up to it? Somebody got to do it… and it will eventually be a multi-bilion dollar market place… I have no doubt about it.

  12. Sell storage space as a ongoing fee for service, collect a percentage or all apps purchased by the client, sell premium access to drug companies to a annonymized database of causal variations 🙂 … result 🙂

  13. BaseSpace 😉

  14. In principle you are right Mick. In practice there are still a few hurdles – one is the size of the data – though storage will become less and less of a problem as time goes by as we know (though the data has a tendency to increase in size too, as new forms of data are incorporated). A second is security of the “storage” facility – whatever it may be – almost certainly now a good idea to have it on one’s mobile phone. Thirdly, public understanding of the data and uncertainty/risk in general. Nevertheless, any DTC should provide some measure of how confident they are that an observed mutation was correct – do any of them actually do that, and do we know if there is any external validation of whether the figures they present are correct?

    A huge niche in terms of interpretation is waiting to be filled though.

  15. PS. A further issue I forgot to mention is that it isn’t currently easy to get a second opinion on such matters i.e. I give you my FastQs, can you tell me which worrying SNPs I have?

  16. I agree these are important issues, and what I would like to see is some very broad legislation that:

    1) the patient, and only the patient, can own their genomic data
    2) that a patient’s genomic data can only ever be used for the benefit (medical, financial or otherwise) of the patient, never to the detriment.

    I don’t even know if that is legally possible, but that’s what we should aim for.

Leave a Reply

© 2017 Opiniomics

Theme by Anders NorenUp ↑