Selventa Says Systems Diagnostics is Key to Harnessing Patient Data

By Kevin Davies 

January 31, 2013 | To hear any CEO claim that their company has become “the premier big data analytics company focused on personalized healthcare” might raise an eyebrow or two. When the company in question is a small software firm that underwent a name change just 18 months ago, that is certainly the case.

But Selventa CEO David De Graaf says his firm’s long-standing expertise in systems biology and biomarker discovery is now being refined and extended to the patient—by supplying physicians the tools to determine if their patients will respond to a given drug.

“We’re raising funds to start to bring this to market,” he says. “We have significant interest from several investors. We think, to a large degree, this is the future of computational biology companies. We’re at the intersection of the commoditization of big data.”

Over the past year, De Graaf says Selventa has performed a number of successful biomarker discovery partnerships with pharma companies such as Janssen Pharmaceuticals to deliver companion diagnostics -- or as he puts it, “the analytes to distinguish responders and non-responders.” Areas of particular focus include oncology and autoimmune diseases.

Now, De Graaf says his team has rethought and renamed its platform. “We’ve honed down and brought together our core knowledge in systems biology with its application in advanced diagnostics. We’re opening up a new branch of complex diagnostics called systems diagnostics.”

While it raises money, Selventa will be developing a line of tests that will be available to physicians and payors, initially to identify which patients will not respond to the standard of care in rheumatoid arthritis (RA), namely anti-TNFs.

“This was a gleam in our eye last year, but we now have compelling clinical evidence from RA patient biopsies and blood that we can identify non-responders with great accuracy,” says De Graaf.

Pharma Path 

The new push into systems diagnostics is but the latest iteration for Selventa, which until 18 months ago was named Genstruct.

Before joining Genstruct in 2010, De Graaf has been one of biopharma’s leading evangelists for the field of systems biology. He initially trained as a molecular biologist, creating one of the first retroviral vectors used in gene therapy trials as a graduate student. He worked for a spell with MIT’s Doug Lauffenberger and then with Eric Lander in the late 1990s. “I'm an author on the human genome project paper -- number 250 -- and I'm on the T-shirt,” De Graaf says proudly. “They misspelled my name on that. That's about the level of my contribution there!”

In 2000, he joined AstraZeneca, followed by spells at Pfizer and Boehringer-Ingelheim. He reflects about the industry’s reaction to the patent cliff and desire to invest in innovation and double down on technologies likely to make a difference. “A deeper understanding of basic biology was clearly one of the needs because chemistry was industrialized but the biology never was. The simple question was: can we understand our biology better up front? That was never particularly refined.”

At AstraZeneca, De Graaf helped build a systems biology team, pieces of which are still spread throughout the organization, he says. At Pfizer, the effort to gain a more basic understanding of complex biology focused on three main areas: predictive toxicology, intracellular signaling and pharmacokinetics/pharmacodynamics (PK/PD).

“We worked on 80 biotherapeutics projects and helped them decide which ones had the right approach,” he says. Later, he led the first biotherapeutics group at Boehringer-Ingelheim in the United States, pushing two compounds into the clinic.

PK/PD issues are a major reason for drug development failure across the industry. “Ten or 15 years ago, people thought that doing PK/PD modeling was absolutely useless. And then people started using computational tools and the science has developed and groups have flourished -- it's now a relatively well accepted science.” He says the science was developed before there a clear understanding of the mechanisms behind drug clearance. This led to the emergence of PB/PK (physiologically-based PK modeling) and the launch of companies such as like Pharsight.

While at Pfizer, De Graaf says his group developed much more quantitative data than other teams, collaborating with software start-ups such as Entelos, Gene Network Sciences, as well as Genstruct, a company founded by former Merck executive Keith Elliston. “We were trying to figure out what you would get out of different ways of viewing the data,” he says.

Early on, Genstruct focused on the area of predictive toxicology. It was a great field to apply novel strategis, says De Graaf, but “a hard place to monetize from the perspective of a small company.” Nevertheless, that was his next stop: in May 2010, he joined Genstruct as chief scientific officer.

Name Change 

When De Graaf joined Genstruct, the company was essentially doing scientific consulting, but extracting full value from the biological insights was left to the customer. “We went through 60 projects that the company had done with various partners and asked, ‘Where do we bring unique value?’ It was clear that our unique capabilities were in understanding disease drivers in single patients, which is very different from the idea of virtual patients or combining data from a large number of different places. We felt there was a major need in personalized medicine to understand how one patient differs from the other in terms of disease-driving mechanisms.”

As Elliston left the company in 2011, Genstruct’s business model was evolving from a typical fee-for-service consulting firm to applying the platform to help analyze individual patients and create biomarkers to prospectively stratify populations. “By the time we understood where we wanted to go, I was raring to go and threw my hat in the ring [for CEO],” says De Graaf.

The company’s pitch to customers was: ‘We have these unique capabilities. Where are your issues and your problems? Give us the data and we'll help you move along.’ That's a very powerful way to develop your own internal platform, De Graaf says, but there were problems with that approach, including transparency. “Once we understood that what we were delivering are biomarkers for patient stratification, it became much less important to hold onto the platform itself.”

Last year, De Graaf elected to release some of those formerly proprietary tools as well as the representation of the underlying biological data—OpenBEL. “We think that everybody should be building on this,” he says. “We've worked on it for ten years, and now we want to share. It provides more value if we work on this together.”

As for the name change, De Graaf says the Genstruct name was about technology, while he wanted to signal new concepts of openness and clarity. One of his colleagues typed ‘clarify’ into Google Translate. “It turns out that in Finnish, the word for clarify is Selventa. We thought that sounded really good. It's a great name!”

Selventa quickly targeted autoimmune diseases a particularly promising field. After extensive research, De Graaf concluded that Selventa could do “a decent job a priori of saying which patients will respond to a major drug class -- anti-TNFs.” These drugs account for about $18 billion/year out of the $24-billion RA market. But for that sum, only half the patients actually benefit from the first-line therapy.

“If there's one thing I've learned over the last three years, if you don't start with the patient but with technology, you're always going to be in search [mode]. Where you end up may not be all that valuable. So we’re starting with the patient… We're switching from a technology-based view of the world to an applied view of the world.”

Selventa has focused its biomarker discovery efforts in gene expression and proteomics. With some clients, the company analyzes secreted proteins detected in blood, and launched clinical trials to identify patients more likely to respond to a given therapy in oncology. With others, the focus is on distilling gene expression data to follow as a panel to help decide the patient is likely to respond.

Six months ago, as De Graaf was evolving his thinking on market strategy, he told Bio-IT World: “My strong opinion is that there is a way in which patients are already being approached by doctors. We need to make use of that information. We can't discount the fact that before you put a fancy test in between a physician and a patient, first they're going to look at age and general clinical presentation. They'll do blood work that we need to integrate one way or another. How do you use what's already been collected? Then if you have to do anything fancy on top of that, we're ready for it and we can integrate it.”

Payors and Physicians 

Seleventa is launching a decision-support tool to identify RA patients that won’t respond to first-line therapy, anti-TNFs. “This test reports sits on top of big data generated on each individual patient,” says De Graaf. “Physicians and payors are interested in identifying patients that won’t respond to standard of care drugs. Otherwise, costs go up, driving bad economics in today’s healthcare system. They want to know if patients respond, and if not, let’s move them onto something else.”

Selventa’s new blood test will be processed and analyzed for gene expression and proteomics profiling. Company scientists will analyze the information. The ensuing report, delivered to the physician, will inform patients if they are likely respond to anti-TNFs. If not, they will have the eability to select from among many alternative available therapies, thus sparing them a diagnostic odyssey of sorts.

De Graaf says there are some parallels to the model Foundation Medicine is pursuing in cancer—generate “big data” on a patient, analyze the information, deliver the individualized report to the physician. But the real value there, he says, is not the low cost of sequencing but in the consistent interpretation of that piece of information to drive treatment decisions. “The holistic integration of data is something we’ve specialized in for many years. We can bring that feature to many more patients now,” he says.

What are De Graaf’s views on big data? “We’re not talking about indiscriminately deploying every device for all patients and having a magic machine spit out an answer! There’s evidence for patients in EMRs on the molecular basis of disease and we have ability to understand the molecular read outs of these patients.”

De Graaf says there are a handful of companies doing innovative work that parallels and complements Selventa’s efforts. Crescendo Biosciences, a spin-out from Genomic Health, is one. “They used some smart sleuthing to get a panel of 12 proteins and algorithms to allow an RA patient to ‘score’ their disease and receive an objective measure of disease progression… They did the right clinical research and filtered the right clinical measures.”

He also cites Exagen, Agendia, and Vermillion as other companies innovating in this area. “We don’t want to be in business of generating the data or rebuild it. We want to have a scalable platform across a variety of diseases.” From RA, De Graaf has his sights set on inflammatory bowel disease and oncology.

Lest anyone think De Graaf is getting carried away with lofty predictions around predicting response in personal health, he states: “We have a narrow niche. We’re not doing prognostic biomarkers or cancer recurrence. It’s about whether the choice the physician prescribes is the right choice.”