Skip to main content
Premium Trial:

Request an Annual Quote

Q&A: Decode's Stefánsson on Using Whole-Genome Sequencing to Link Rare Variants to Common Disease


KarisStefanssonPhotoSmall.jpgName: Kári Stefánsson
Age: 62
Title: Co-founder, chairman, and CEO, Decode Genetics, since 1996
Experience and Education:
Professor of neurology, neuropathy, and neuroscience, Harvard University, 1993-1997
Director of neuropathology, Beth Israel Hospital, Boston, 1993-1996
Assistant, associate, and full professor, departments of neurology and pathology, University of Chicago, 1983-1993
MD, University of Iceland, 1976

After publishing scores of genome-wide association studies that used array-based SNP genotyping to identify common variants that increase the risk for common diseases, Decode Genetics recently shifted its focus to discovering rare variants that contribute to these diseases.

The company has been using a strategy that starts with whole-genome sequencing of several thousand Icelanders to discover rare variants and then imputes these variants in a larger fraction of the Icelandic population who have been genotyped by arrays. So far this year, the company has published four rare risk variants linked to ovarian cancer, sick sinus syndrome, gout, and a number of cancers. Decode has made additional discoveries that it plans to publish in the near future. It also recently forged an alliance with Pfizer to discover risk variants associated with lupus erythematosus and expects to add other such partnerships.

Clinical Sequencing News recently spoke with Decode co-founder and CEO Kári Stefánsson about the company's approach, how it plans to translate its findings into new tests for disease risk, and when customers of DecodeMe's direct-to-consumer offering may benefit from the results. Below is an edited version of the conversation.

Tell me about Decode's whole-genome sequencing project and its goals.

The tidal wave of discoveries that came out of genotyping with SNP chips that capture common variants — we were responsible for a very significant proportion of those — left a substantial amount of heritability unaccounted for.

It was absolutely clear it would be very difficult to ignore the rest [of the variants]. And even though we could use haplotypes made up of common alleles — if you string them together, you can end up with rare haplotypes that can tag rare SNPs — in the end, we would have to sequence to get the rare SNPs. So [by doing whole-genome sequencing], we're basically leapfrogging over the necessity of doing deep sequencing of particular regions.

As soon as it became economically feasible [to do whole-genome sequencing], it was the way to go, no question about it. So we went ahead and started to do this, and we have been very successful in using this to make discoveries.

The reason that we are in a very good position to do so is that we can impute rare variants in Iceland that people cannot do elsewhere because of the population structure that we know in detail. Our plan is to sequence the whole genomes of about 3,750 Icelanders. We are already done with about 1,400 to 1,500.

We're doing this with Illumina HiSeq machines; we have currently 14 of them in our facility. The coverage is between 10X and 30X. We call variants in these samples — we are now up to about 22 million SNPs. [We also perform] genotyping with an Illumina SNP chip on a total number of about 120,000 [additional] people. That allows us to impute the whole genome sequence, down to a variant frequency of about 0.1 percent.

Why can we do that? Because we can take any two randomized samples and we can pick out what they have inherited from a common ancestor. For each million-base set, we can find about 200 to 300 Icelanders who have inherited these from a common ancestor, and we only need to sequence [this set] in one of them to be able to infer the sequence of the others. Once we have done this in a sufficient number of people, we have basically nailed down the population structure of the entire nation.

If we were to continue to sequence at even greater depth, we would probably be able to find even rarer variants, but what we have now is about 22 million sequence variants.

In addition to this, we enjoy the benefit of having a founder effect for many traits in Iceland.

For example, we recently published a paper on ovarian cancer in Nature Genetics, where we took advantage of a founder effect in the Icelandic population. We discovered a founding mutation that has an allelic frequency of about 0.4 percent in Iceland, and another founding mutation in the same gene in Spain, replicating our finding.

We discovered another founding mutation for sick sinus syndrome that we published in March, one in p53 [that confers susceptibility to several cancers], and one in aldehyde dehydrogenase that is associated with gout.

We have similar variants in about 20 other diseases that we have already discovered.
Once we have found them in Iceland because of the founder mutations, we go into outbred populations to map out the mutational diversity that you find there.

So even though these are founding mutations in the Icelandic population, they can still tell you about disease risk variants in other populations?

The mutations are just much more frequent in Iceland. You also find them, but at a much lower frequency, in other populations, and then you find many other mutations, just very rare mutations, in these populations.

Let me give you an example. About 20 years ago, the BRCA2 gene was discovered that confers very high risk for breast cancer. It was discovered in two places simultaneously, in Utah and in England, and it was discovered because both groups had Icelandic material. They found it because there was a founding mutation in BRCA2 in Iceland.

Myriad has now been sequencing this gene as a test for many years, and they have found over 6,000 mutations in the BRCA2 gene in America. The allelic frequency of the one mutation in Iceland is about 0.4 percent. The combined allelic frequency of these 6,000 mutations in America is about 0.1 percent. So you can see that it's almost impossible to find these mutations in America, but once you have found them in Iceland, you can map out the mutational diversity in America, so that the founder population in Iceland is absolutely golden as a discovery population. It allows us, then, to go out and find the additional variants in outbred populations.

The Icelandic individuals whose genomes you are sequencing are part of a group of 140,000 volunteers Decode recruited earlier — did you have to re-consent them for this?

We had consented them for mapping out the diversity of variants in the genome, irrespective of the methods.

How is Decode going to translate the findings from its whole-genome sequencing studies into new products?

If you look at our ovarian cancer paper, the mutation that we found that confers risk of ovarian cancer is very similar, and the gene is very similar, to the BRCA genes. Therefore, you can basically see the clinical utility of this discovery being reflected in the clinical utility of the BRCA tests. So it's an example of a discovery that is public, and that we can translate very quickly into a clinically relevant test.

Do you want to develop the ovarian test yourselves at Decode?

The ovarian cancer mutation has all of the qualities of a discovery that we might want to take ourselves to the market. There is a lot of unmet need — the five-year survival of patients with ovarian cancer is only about 45 percent. It is extraordinarily important to find [high-risk mutations] because it has been shown that if you take women at very high risk of ovarian cancer, you can prevent death from the cancer, so there is a compelling need for a test like that.

What is the timeline for developing this test?

I cannot give you an exact timeline because the regulatory authorities are trying to figure out how to regulate a test like this. My guess is that it's not going to take much more than two years to get such a test on the market.

How many tests do you think you will be able to develop from your discoveries, and in what areas?

The discoveries are not going to be the limiting factor, but our ability to convert the discoveries into tests. We are making so many of these discoveries of [rare] variants with very large effects, so we will have to be very selective which ones we take ourselves to the market. We will probably outlicense many of them to other diagnostic companies.

Can you talk about Decode's recent alliance with Pfizer (CSN 10/12/2011)?

We're working with them on systemic lupus erythematosus. Lupus is not a particularly common disease, it's a relatively rare disease, [but] it's a very familial disease with a large heritability. [For this collaboration], we are sequencing a formidable group of Icelandic patients with lupus at greater depth than [we are doing in our own project]. The project has been ongoing for a while; we are making good progress.

Our role is to discover genes with variants that confer risk for the disease. Variants in the sequence do not confer risk in a mystical fashion; they do so by upregulating or downregulating biochemical pathways. So the goal of a pharmaceutical company is to put together compounds that contain the pathway that is perturbed. We are working with Pfizer to find genes in biochemical pathways to give them targets to work with.

Where is the funding for Decode's research currently coming from?

Saga Investments [an investment consortium with financial backing from Polaris Venture Partners and Arch Venture Partners] is funding some of this, and we have significant money in grants, both from the National Institutes of Health and from the European Union. And now we have substantial revenue coming from our alliance with Pfizer, and we are expecting revenue from other alliances that we will announce in the weeks to come.

You also have a direct-to-consumer business, DecodeMe. 23andMe recently introduced an exome sequencing pilot program for its customers (CSN 10/5/2011). What are DecodeMe's plans in that area?

To introduce exome sequencing now as a service for consumers is alright, but it is a little bit deceiving because exome sequencing allows you to find rare variants, but there are [currently] very few rare variants that associate with common complex traits. Basically, the only ones that have been published are the ones that we have [recently] published.

What you are offering by offering exome sequencing now is an opportunity for people to have the sequence, so they can wait and see what happens when we begin to make discoveries of the variants. The problem with that idea is that the price of sequencing is coming down very fast.

We are not going to be offering exome sequencing or whole-genome sequencing to consumers until sufficient discoveries are made to make them useful to the consumer.

How long will that take?

Probably within a year or two, it will become useful, but not until then. In spite of all of the effort in whole-genome and exome sequencing that is taking place these days, there is not all that much coming out yet. But that will change.

I would advise anyone who aspires to have their exome sequenced in the near future, just wait until we have something that makes it valuable, and then the price of having this done is going to be a fraction of what it is [today].

How will this information be valuable in the future?

It's going to be valuable like any other risk assessment. If you believe that it is going to be of value for you to know what risk you have of developing common diseases, then this is going to be the best way of having your risk assessed, because at least about 70 percent of the risk of all of common diseases is genetic.

So if you believe in preventive medicine and personalized medicine, there is going to be enormous value in it, but the value is going to be proportional to how much of the genetic risk has been unearthed.

And my expectation is that we are going to see an explosion of discoveries in that field over the next year or two.