Amgen Mulls Future of deCODE’s Clinical Sequence Miner Platform

By Kevin Davies 

March 29, 2013 | BOSTON—An interview with Kari Stefansson, the co-founder, CEO, heart and soul of deCODE Genetics, is an experience quite unlike any other biotech executive briefing.

One moment he is kissing the hand of the young hostess at his luxury Boston hotel where we’ve arranged to meet, the next he is joking (I think) about lopping off the head of a potential client he’d met earlier in the day. At one point, he abruptly halts our discussion to speed dial a colleague, whom he promptly berates in a torrent of Icelandic expletives for a trivial misdemeanor—before telling his associate it was all in jest.

Stefansson agreed to an interview during a pit stop in Boston late year, en route to Los Angeles for a trip that was part business, part pleasure. His two daughters reside in LA (one of them married the son of ex-Beatle George Harrison last summer), and he casually mentions that in 24 hours, he’ll be dining with one Bob Dylan.

On the business front, he also mentions that he would be paying a visit to Amgen. The significance of that remark flew completely over my head, only becoming clear a few weeks later, when Amgen announced that it was acquiring the Icelandic firm for a cool $415 million.

Most of the analysis regarding Amgen’s acquisition has focused on the putative benefit of leveraging deCODE’s genomics prowess on boosting Amgen’s drug development pipeline. Although the price tag looks steep, it wouldn’t take too many successful drug development programs over the next few years for the deal to pay for itself.

Sasha Kamb, Amgen Senior Vice President Research—to whom Stefansson now reports—opts for a baseball analogy. “We’d recoup our investment if it turns out that this player we traded for is a power hitter or hits for average. But it’s possible that deCODE can do both,” he told Bio-IT World.

“Because of the costs of clinical studies and [developing] successful drugs, you don’t need to have a lot of hits to justify [a deal] like this. But we’re also hopeful that [deCODE] will hit for average—that there will be many smaller, earlier effects—information that will give us the conviction to aggressively move [drugs] forward or aggressively terminate things.”

As Kamb and colleagues assess their newly acquired assets, they will find some other intriguing possibilities. One in particular will be the fate of the Clinical Sequence Miner, an ambitious clinical knowledgebase that deCODE outlined in public shortly before news of the Amgen deal broke.

Road to Recovery 

Even before Amgen’s overtures, 2012 was shaping up as a very good year for deCODE. A series of high-profile publications in Nature and the New England Journal identifying new targets for Alzheimer’s disease and a plausible mechanism for the rising prevalence of behavioral disorders such as autism, was nothing new for deCODE. For all its boardroom troubles, the Icelandic firm has routinely captured headlines for a decade as Stefansson and colleagues mined the trove of genetic and medical information from the Icelandic population for novel insights into population genetics and disease risk.

But after swelling to 750 employees and a market cap exceeding $1 billion, an economic perfect storm—deCODE’s expensive ventures in drug discovery, Iceland’s economic collapse, and more specifically the loss of capital reserves held by Lehman Brothers—decimated deCODE.

Stefansson admitted that deCODE’s earlier foray into the drug development business was “an attempt to pull a rabbit out of the hat” at a time when genetics had not matured to the point that deCODE could systematically provide genes for diagnostic tests. He insisted that the three drug programs deCODE launched—two in cardiovascular disease, one an anti-platelet therapeutic—were and remain “very interesting compounds.” At least one of them is being picked up, and he said he would be surprised if they weren’t eventually brought to market.

“But that’s not our expertise,” he said. “We’re geneticists.”

The entity that emerged from the 2009-10 bankruptcy period was a much leaner, more focused entity. But asked what he learned from the bankruptcy process, Stefansson simply shrugs: “Nothing at all. We started the company eight years too early. But we wouldn’t be the powerhouse we are today if we hadn’t started that early. We wouldn’t have the database we have… The kind of genetics we were doing has been too pre-competitive, too difficult to attach it to something money-making. But it’s becoming easier now.”

Asked to define the new post-bankruptcy deCODE, Stefansson paused briefly, perhaps in retrospect a sign that the definition was in flux.

“We have one core competence, which is in human genetics,” he said. “We’re a small operation, about 125 individuals, but despite our small size, we’re probably the most productive operation in human genetics—and have been for some time. We have an extraordinary amount of data and analytical systems and very good scientists in the clinical sciences.

“But we don’t have anything more than that. We don’t have a good sales organization. We’re building up a business model in clinical sequencing, which is based on helping others to generate revenue, wherever it may be—hospitals, biotech companies, etc. And I think we’re going to succeed, because there is so much need.”

Clinical Sequence Miner 

It was in an effort to address that clinical need—and leverage deCODE’s strength in personnel and discovery data—that spurred the development of a software platform called the Clinical Sequence Miner (CSM).

After 15 years of discovery research, Stefansson said deCODE was perfectly positioned to move into the clinical arena. Whether discovery or clinical care, “You use exactly the same software systems and analytical algorithms and know-how when you do these two things,” he said. “We have whole-genome sequence on 70,000 Icelanders, so we’re in an incredibly good position.”

“I know this sounds a bit arrogant,” Stefansson continued, “but there are not many places that have anything that comes close to this.”

Many groups and start-ups have been putting together systems in a vacuum without the data, said Stefansson, who conceded that some might eventually become good systems. “But before they have run them on a substantial amount of data, they will have to run through an awful lot of problems we have already solved.”

The CSM—a comprehensive clinical data IT infrastructure and interpretation suite—is “tailor designed to clinical problems. Our discovery work has all been clinical science—we’ve been trying to shed light on clinical disorders,” he says. “Our software system… has been tried on an enormous amount of data, and it works. That’s all there is to say about it.”

Leading the development of CSM has been Hakon Gjubartdsson, who outlined the project at the Bio-IT World Europe conference in Vienna last October. Gjubartdsson is a close friend of the deCODE founder; his father was Stefansson’s high school geography teacher, and his wife taught Stefansson’s daughter the flute when they were colleagues at Harvard Medical School just before launching deCODE.

deCODE was in the early stages of promoting CSM to various hospitals and clinical groups when the Amgen deal closed.

According to company documents, the CSM software is designed to integrate next-gen sequencing into clinical practice using “state-of-the-art software solutions for research in genetic analysis at petabyte scales.” The browser-based system allows patient data to be pulled from a third-party electronic health record, rather than try to accommodate sequence variant data in an ill-suited legacy system, and integrated with genome sequence data.

Using the CSM, for each patient a clinician can select terms representing symptoms from the Human Phenotype Ontology, helping select candidate genes for analysis and enabling systematic comparison of variants from cases and controls. The CSM system picks up the patient file containing the sequence data, either delivered via portable disc drives to deCODE or via an ftp connection to a staging folder in the cloud.

For each genome sample, the analysis pipeline performs a thorough annotation of the sequence, outputting aligned sequence reads, a VCF variation file, as well as annotated variations and sequence coverage data in a genomic ordered relational (GOR) file format.

The CSM provides four major categories of reports: 1) Sequencing quality, 2) candidate genes, 3) all known variations, 4) putative pathogenic variants. Before a suspected pathogenic mutation is used for patient diagnosis, the clinical and bioinformatics team members can review and interrogate the raw sequence read data, retrieved using a synchronized Java Webstart genome browser. Another feature of the CSM is the ability to add textual comments and/or labels from a menu for any interesting variants. From here, a customizable summary report is printed. DeCODE says the CSM was designed to “support collaborative workflows, allowing physicians with only moderate skills in genetics to investigate patients in collaboration with experts.”

Do No Harm 

Building powerful clinical informatics pipelines is a major challenge in the rapidly growing field of genomic medicine. As Amgen decodes its new Icelandic assets, Kamb and colleagues must now decide whether this is something worth potentially developing or spinning out. Prior to the Amgen deal, Stefansson was talking to potential clients about “revenue-generating operations” featuring the CSM.

Kamb admits that Amgen didn’t acquire deCODE because of its CSM software. “It’s a small part of the equation,” he says, “but it does look impressive.” That said, “The whole IT Infrastructure they’ve built is really impressive.”

As a general principle, Kamb says, “We want to make sure that anything that is useful to people gets out there.” There is a precedent for Amgen spinning out a new company: last year it launched Atara Biotherapeutics with a venture capital firm, but Kamb would not be drawn on whether the CSM asset could similarly be spun out.

The deal to acquire deCODE happened quickly. Kamb first met Stefansson some 18 months ago. “We wanted to proceed slowly, but it became clear they’d really embraced the technology revolution around ultra-high-throughput sequencing and genotyping, and Kari’s vision was becoming reality. There was a real opportunity to do something much more aggressively.”

Kamb’s colleagues quickly agreed. “No-one had to convince anyone else. It was quite amazing,” he says. Kamb, who led the team that isolated the BRCA1 breast cancer gene in 1994 while working at Myriad Genetics, says he developed “a good personal connection” with fellow gene-hunter Stefansson, whom he praises for having “visionary qualities.”

It will be a while before the wisdom of the deCODE deal can be judged. “We’re in the profession of trying to perturb human biology in positive ways and helping very sick people,” says Kamb. “That runs the gamut from innovative new therapies to understanding how to use the therapeutics you are developing or have already developed, both on efficacy side as well as safety and tolerability… In my world, it’s mostly about innovative new drugs. It’s about picking the winners.”

Kamb says Amgen has no intention of messing with deCODE’s scientific success. “We’re trying to tread pretty carefully,” he says. “They have this fantastic engine and they’re cranking out all kinds of stuff. We’re going to help them with a drug hunter’s intuition.”

Amgen isn’t a diagnostics company, and a company spokesperson said there are no plans to continue offering the deCODEme personal genomics test to the public. But the deCODE acquisition could have other benefits on Amgen. Kamb says: “The rigor [deCODE] will infuse our culture is an intangible that shouldn’t be dismissed either. It will make us all a lot smarter.”