How do you find disease-causing genes? Kari Stefansson thinks he knows the secret.
Kari Stefansson is in a hurry. The president and CEO of deCODE genetics must get across town to see a former colleague at Beth Israel Deaconess hospital before dashing to the airport-and he still needs to squeeze in a promised interview with a journalist. In the cab, he alternates between discussing travel plans in Icelandic with a coworker and expounding on the intersection of disease and human evolution in English with TR associate editor Erika Jonietz.
Changing how genetics is done has kept Stefansson on this hectic pace since 1996, when he left Harvard Medical School and Beth Israel to found deCODE genetics, based in his native Iceland. In 1997 the company proposed construction of the first “phenotype database,” a collection of the health records of all 280,000 Icelanders. DeCODE would use information from this centralized health-care database and Iceland’s extensive public genealogy records to find disease-causing genes, aided by the relatively homogenous genetics of Iceland’s population. This plan helped launch a revolution in population genomics (see “Your Genetic Destiny for Sale”), with a variety of other companies quickly following suit.
Although it’s widely imitated, deCODE remains controversial. Editorials in publications from the New York Times to the New England Journal of Medicine attacked and defended the ethics and scientific merits of the proposal. After extensive public debate, Iceland’s parliament approved the creation of the Icelandic Health Sector Database in December 1998 and granted deCODE a 12-year license to create and operate the database last January. A vocal minority, including the physician-led group Mannvernd, is still trying to stop the database in the courts, but deCODE has moved ahead. Last July, the company completed a successful initial public offering that raised $244 million.
TR: You were a professor at a leading medical school. Why leave and switch directions so radically to found deCODE?
STEFANSSON: Basically, the methods of genetics are the methods that you use no matter the disease you are studying. Around the time I founded deCODE, the technology had evolved to the extent that it had stopped being the limiting resource; the limiting resource was becoming the access to populations to do this work. So I moved to a place where there were the least limitations on that resource. I think the Icelandic population is not unique, but it’s very well suited to do exactly what we want to do, which is to apply this new technology to a well-defined population.
TR: What has it been like for you to be a CEO instead of an academic scientist?
STEFANSSON: The difference between the two is vastly overrated. What mostly drives you is the desire to win, to perform, to control your own fate, whether you do it through money, or the admiration of people who follow your work or whatever. It’s a larger scale: I was running a lab of 10 or 15 people; I’m now running a company of about 450 people. But it is basically the same thing. You put together certain ideas, you gather around you people to execute them. The fact that you can create value out of the results of your research doesn’t alter in any way the weight or the importance of the fundamental question you are asking.
Also, don’t forget that I come from Iceland. My family has lived in Iceland for 1100 years. There is a certain adaptation that has taken place. We fit this sort of wet, barren, dark corner of the North Atlantic. That does not necessarily mean I like every aspect of it, and I miss America a lot. It was a great place for me; this was a community that was extremely generous to me. I learnt an awful lot here. I’m running a company in Iceland that is fundamentally run on American philosophy. It fits pretty well there.
TR: What specifically motivated you to create a phenotype database?
STEFANSSON: When you look at common diseases, they are probably common because they are complex, and they’re complex for all kinds of reasons. Not only because they probably require the confluence of many genes to cause the disease, but they may also require interactions between genes and the environment. And when you are studying complex common diseases, you have to work on the assumption that you know very little. In spite of 75 to 100 years of intense research into these diseases and an enormous expenditure of money, we have come a short distance in developing a useful understanding of these diseases, simply because the fundamental approach just doesn’t reach deep enough. You don’t see enough of a pattern to put together a reasonable hypothesis. You basically have to begin to use systematic data mining, to be able to bring together large amounts of information.
TR: How does a phenotype database help you find the genes behind diseases such as asthma, heart disease or diabetes?
STEFANSSON: You may have someone with diabetes in your grandfather’s generation and it skips the parents and then it affects the children, and so on. And to study these disorders, you have to go to a population, because once a disease begins to skip generations, a nuclear family isn’t a useful unit of society to study.
We can design experiments to mine knowledge out of this [population] data by putting together hypotheses, the old intuitive approach. But as beautiful as that approach is, it is not very powerful. The alternative is to use combinatorial analysis, to take every single data point and compare it to every other data point, looking for the best fit, unblinded by a hypothesis. The unaided human mind cannot handle large numbers of data points; you need the modern computer. It’s an extremely powerful approach. And I’m absolutely convinced that we will revolutionize the ability to develop knowledge about the common diseases by using systematic data mining.
TR: What is deCODE’s advantage in this field?
STEFANSSON: We have the genealogy of the entire nation [of Iceland] on a computer database. When you’re studying the genetics of disease, you’re not only studying the information that goes into making an individual; you’re also studying the flow of this information from one generation to the next. And the genealogy gives you the avenues by which the information flows.
TR: Have you had any success yet?
STEFANSSON: We have indeed had success. We have been able to map genes in incredibly complex diseases, like osteoarthritis, osteoporosis, schizophrenia and psoriasis. Every single complex disease that we have applied significant effort to, we have been able to map by using this process. When we were starting our company, we put most of our effort into development of software systems, of algorithms to analyze data, and we were ahead of the biology that we subsequently began to look at. Making use of informatics, mathematics, and all these genealogical approaches, we have been able to solve problems that others have had difficulties with. Then we are taking the discovered genes and turning them into drug targets. And we are already working on the development of drugs on the basis of those targets.
TR: What kinds of drugs is deCODE working on?
STEFANSSON: We have set up a division to work on drug development, which will probably be located in the States for human resources reasons. It started out as a cell biology division where we work on the biology of the genes we discover. Since we are working on the genetics of 60 diseases now, we have the possibility of selecting the things that are easiest to work on. I honestly cannot tell you which genes we plan to target, because we haven’t made any final decisions. Remember, if you are speaking in terms of drug development based on basic biological discoveries, it takes about 12 to 15 years from the time you make a discovery until you have a drug on the shelves.
TR: How does all this fit into deCODE’s business model?
STEFANSSON: We are basically marketing three lines of business: a discovery service, from discovery of genes to the discovery of drug targets to the discovery of drugs; the service component, which is the database that is based on putting health-care information in the context of genetics; and software systems, not only to use in discovery but also in the delivery of health care. When we started our work and we looked around, there were no software systems that would allow us to do genetics at the scale that we wanted to do it, so we put together a very large informatics section. They have put together algorithms and programs that do spectacular things when it comes to large-scale genetics, for genotyping [finding the individual variations in genetic makeup], for mining genes from raw sequences, for doing statistical analysis.
We have also put together a software system to prevent medication errors, problems due to drug interactions and things of that sort-a very smart, very elegant program. We have put together software systems that protect privacy not only in discovery work but also in the delivery of health care. Now, and particularly in the future, people will be sending medical information to institutions left and right in electronic form. To do that, you need to put a shell around the data so you can protect privacy. We have also been putting together a clinical decision-support system. All of this grew out of our focus on the database.
TR: Who are the customers for your various products?
STEFANSSON: The discovery stuff goes mainly to pharmaceutical companies and diagnostics companies. The database services, to academic and other kinds of health-care institutions. The software systems we are marketing to health-care providers, the pharmaceutical industry, biotech companies, software companies. The breadth of customer is considerable. This is not just biotech companies serving pharmaceutical companies.
TR: There has been a lot of controversy surrounding deCODE’s creation and licensing of a database containing the health records of every Icelander.
STEFANSSON: The database has been controversial mostly for the wrong reasons. There are all kinds of reasons to be skeptical of collection of personal information, and I think that we can never be too careful when we do that. But most of the controversy was focused on misinformation, the insistence that we were working on biological samples without informed consent and things of that sort.
TR: How do you reassure the public of the value of these databases?
STEFANSSON: There is no question in my mind, nor in the mind of anyone who has looked at this carefully, that this is an extraordinarily important approach, and people should take it. The only question is, What is the price? What are the sort of ethical and societal dilemmas that you have to overcome to be able to take this approach? If you’re going to do a large study that involves a large number of people, a whole population, you have to establish some sort of a consensus in the population whether it should be done-what some people call community consent. In Iceland, people took my suggestion to do this sufficiently seriously that the parliament passed a law. We could easily have done it without a law, but the parliament passed a law. There was a societal debate that lasted for nine months. There were 700 articles written in the country’s three newspapers; there were 140 television and radio programs addressing this. It was probably the most debated issue in the history of our republic, and on the eve of the parliamentary vote, a poll showed that 75 percent of people supported the bill, 25 percent were against. And now, under one and a half years after the bill was passed, there was another poll taken showing that the support had risen to 91 percent.
But societal or community consent is not enough. You have to have some sort of consent at the level of the individual: What is the “how” of the collection of this data? Does it comply with current practice in the use of health-care information? By law in Iceland, the information in this database is only information produced in the process of delivering health care, nothing else. Permission to cross-reference it with information from DNA is entirely dependent on an explicit informed consent from those who have given us DNA.
TR: How does this differ from the way information is usually collected for medical research?
STEFANSSON: Health-care information is going to be collected with presumed consent [in which permission is assumed and must be specifically refused], and there’s not a single place in the world where people use information produced in the process of delivering health care with anything except presumed consent. There is not a single significant study in your country where people have demanded informed consent for the use of information produced in the process of delivering health care. So this is exactly in keeping with that, and before you decide anything else, you should think about the consequences of demanding informed consent for secondary use of health-care information. It’s a serious decision, because you would not be able to do epidemiology as we do it today. If it had been required in the past, we wouldn’t have the health-care system that we have today; there’s no question about it. When you come today to seek health care, you’re using the consequence of the fact that your parents and their parents allowed the use of their health-care information. Simply to discover knowledge. And if you refuse to do the same, you’re going to diminish the probability that your children and their children will have health care of the same quality.
TR: How do you collect the genetic information you need to cross-reference with the health-care data so that you can actually map genes?
STEFANSSON: We have to get informed consent from people to give their blood, to isolate the DNA, to genotype and then cross-reference that with health-care information. We have so far been collecting DNA simply on the basis of individual diseases we have been studying, and we have DNA from about 50,000 Icelanders. We may go ahead and collect DNA systematically, put an ad in the newspaper and ask people to give blood for this purpose. And actually that would be the least invasive method to obtain DNA. When you’re doing it through physicians who take care of patients and relatives of patients, there is a certain coercion involved because you’re approaching people at a moment of at least perceived need. But eventually, because everyone has a disease and everyone has someone in their family with disease, we would have everything we need to do this systematically, even if we would simply approach them on the basis of diseases.
TR: It would seem risky to have the medical information of every person in the country collected in a central database. For example, people in the United States tend to worry about insurance companies using such information to deny them health-care coverage. How do you take advantage of the potential that lies in this systematic data mining without causing too much danger?
STEFANSSON: I have visited about 25 countries this year, and in every country I come to, I take out my ATM card and withdraw money. I can do that because the entire world is a network of centralized databases of personal information on finance. And the only restriction on my access to your bank account lies in this little card I have. It is much easier to abuse personal information in finance than in health care. You actually have to have a fairly lively imagination to figure out how you can abuse it [in health care]. But the reason we have electronic banking is that we feel it provides such comfort to us that we’re willing to compromise on protection. I’m absolutely convinced that less than one percent of the people in this world would believe that the use of electronic banking is a more noble goal than discovering new knowledge in medicine.
TR: How would you address people’s concerns about this, though?
STEFANSSON: The way to deal with any risk from knowledge is not to forbid the discovery of the knowledge, or the gathering or the storage of information; it is to legislate or regulate how you use it. There are so many reasons to have centralized databases in health care. The one in Iceland can only be used for discovery, and that is unfortunate, because if you have centralized databases in health care, you can provide so much better health care. A colleague of mine, his car hit a tree a couple of years ago and he was seriously wounded, he almost died. When he was brought into the hospital, after he had been peeled out of his car, the hospital knew nothing about him except his name. Imagine if they would simply have been able to put his name into a centralized database and get all the health information there was on him, his parents, his siblings. The power of this is enormous, and are you going to say that we should deny ourselves the opportunities that this brings with it because there may be kooks out there who would want to abuse this? The only kooks that could abuse this would be the kooks in the insurance industry. We can simply regulate that.