INSIDE TRACK: Conversation with SVBio’s Dietrich Stephan

Dietrich A. Stephan, PhD

Illuminating the blind spot in clinical medicine

BY KURT WOOCK

          Dietrich A. Stephan, PhD, founder, president, and CEO of SVBio, Foster City, Calif, intends to change the world of diagnostics. He has a big idea and a background of working on big projects. To meet the scope of his vision, he has partnered with Sequoia Capital, the venture capital firm that has helped build Apple, Google, LinkedIn, Oracle, and Cisco. Stephan recently shared his plan with CLP and why he thinks that going big is the way to go.

CLP: Let’s start with the basics. Can you introduce your company, SVBio?
Stephan: So SVBio is a diagnostics company that uses as its input the entire genetic code. As its output, it can provide a genetic contribution to any human disease. That’s our vision. If you zoom out, every human disease has a genetic component. We’ve been testing largely for the 5% of disease caused by Mendelian diseases, single-gene diseases, but there’s a whole other slice of that genetic pie that makes up the vast bulk of disease that is a blind spot. We hope that we can actually improve outcomes. We’re not there today: The company is only about 2 years old.

CLP: It seems like you are building something for a future state of health care. Just what do you see that future looking like?
Stephan: We believe in a future where the genome is very cheap, accurate, and pervasive. For example, instead of a heel prick at birth, a baby would get their genome sequenced. The natural place where that information would reside is the electronic medical record, where the rest of the data lies. We’ve already seen federal incentives subsidizing the adoption of EMRs that make us really believe physicians will be walking around with tablets at the point of care. We believe that in 10, 20, 30 years, the genome will simply be resident at the point of care, with the physician. The problem is that it’s 6 billion letters long. Unless you have a solution that can unmask instantaneously what is significant while the doctor is sitting there with the patient, it will be useless. Our vision is that SVBio is the solution that sits by the doctor and patient and illuminates that big blind spot we have in clinical medicine today, generally speaking.

CLP: You’ve described the challenge and the context. What specific solution will SVBio provide?
Stephan: We don’t build sequencing machines. We don’t build databases. We leverage a lot of public and commercial pieces to build our pipeline. I like to categorize our differentiators at the highest level with a few buzzwords: simple, accurate, and comprehensive.
       First is simple, meaning we allow a physician to input into our system the clinical signs and symptoms, and we will, using experts and a machine, identify the genes TKTKT within the genome and test those. We take a lot of the thinking out of which test should be ordered. To emphasize: If you go to any of these more broad DNA-based diagnostics companies, you’ll see lists and lists of things that can be tested. You really need to be an expert user to make sense of what should be ordered. With a growing number of correlations between the genome and clinical states, it soon will be difficult for those experts to know what to test.
       Second is accurate. Next-generation sequencers and the off-the shelf commercial freeware alignment is geared toward research work, not clinical work. If it wants to perform in the clinical state, it needs to perform as good or better than the gold standard, which is capillary-based. You can’t say, “it’s cheaper for us, and we can test more genes, but a lot more patients are going to slip through the cracks.” That argument doesn’t fly. We actually spent a year trying to get all the protocols for the machines up to snuff. These are preexisting technologies, but there is still a learning curve with running them. Running them consistently is important.
       Third is comprehensive. This is one of the biggest holes in DNA-based diagnostic work. Out of every 100 patients who clearly have a genetic disease, only about 60% get an answer. The other 40% get a report that says, “sorry, no mutations found in the genes we tested.” That is a function of the fact that we aren’t testing in a comprehensive enough manner. We haven’t found all of the genes that contribute to these diseases, or the real estate we are testing isn’t where we [should be] are testing. By testing everything, you have the chance to almost double your sensitivity at a population-wide level. That’s good news.

CLP: What steps do you need to take in order to achieve those three goals?
Stephan: We’ve been quietly building the infrastructure to accept a clinical sample, sequence the genetic code or the relevant portions of it, such as the exome. Next-generation sequencers output billions of very short fragments of a patient’s genome, which then need to be put into a database and TKTKTKT. Thereafter, that genome is compared to a reference genome. The differences are identified. Within any one exome, for example, there may be 100,000 differences between a patient’s genome and the reference genome. The next step in the process is identifying the one or two changes that are causative of the disease that the patient as. That last mile is not trivial. There is a lot of machine learning and algorithmic work and cleaning of databases that happen at that last mile that is really going to fuel adoption of this broad-based assay for clinical work.

CLP: How has your background in the field informed what you’re doing now?
Stephan: When I was young, my mother died of cancer, and my father developed a brain tumor. I grew up with a healthy disrespect of the current state of medicine. After college and my PhD, I went to train at the National Institutes Of Health as part of the national genome project (Human Genome Project?), trying to sequence the first human genome for billions of dollars. I got hooked on this field because you can use that information to very quickly find a broken gene to find a disease and turn it on in a clinical setting to help people. I love that notion of applied research.
       I helped start Navigencis (Navigenics) in 2005. We tried to grapple with the genome as a whole. I think that company was about 5 years too early. Despite having a successful exit, we sold that to Life Technologies. We started SVBio to basically finish the job. Now that price points are within the realm of clinical reimbursement, we said, “let’s start with an exome or genome, let’s start in an area that is getting reimbursed, and let’s drive this next generation of diagnostics in the marketplace.”

CLP: You’ve worked in the private and public sectors. Why do you think your ideas have the best chance to succeed with SVBio?
Stephan: Researchers like me in a past life, generally work on scientifically problems and publish papers and you’re lucky if anyone reads them. And if they do read them the probability of them acting on that information and taking that information forward is sporadic or osmotic in nature. A lot of the research dies on the vine before it ever sees the real world. I was personally disenchanted with that.
       So I decided that if it’s true that my research is aligned with market needs, the best way to ensure that [my ideas are acted upon is to not] I don’t waste my time and my limited amount of energy on things that I can’t control and make sure they see the light of day. Within our mission is the goal to get this into the marketplace.

CLP: The price has come down since that first, billion-dollar genome you worked on. Is it now low enough not to be prohibitive?
Stephan: Today, the price of a clinical grade exome is approaching $750. Again, 23,000 genes covered at hindered-fold coverage. If you juxtapose that against sequencing one gene by capillary-based technologies, it’s almost cost equivalent. We’re probably months away from the inflection point where it’s just as cheap to sequence everything as it is to sequence one gene. The diagnostics are in the software layer. Software is very inexpensive to run once it’s built. By definition it’s largely automated. We can compress the cost of all the interpretation, alignment, into the software layer. If you think about how that might play out in a market price point, you can imagine between a $2,000 and $5,000 genetic test, which is right in line with a monogenic test. You can imagine that you can leverage that information, which is now archived, across that person’s lifetime. You could get even more value for that same price point over time.

CLP: SVBio’s goals are noteworthy. But for folks working in labs of varying sizes, is this going to be a practical solution, or is this going to be science fiction for all but those with significant resources available?
Stephan: Its one assay. One sequence. From a workflow and laboratory management perspective, it’s actually easier to do than dealing with hundreds or thousands of PCR primer sets for various single genes in the genome and running literally hundreds of different workflows through a lab for different genes. Building that diagnostic software is absolutely not trivial: That’s why we’ve built this company. But using it is the true secret sauce. The simplicity of starting with a sample or input genome and simply asking the machine to do all of the interpretive work for you is the difference. Making that simple and easy to use is why we built this company.

CLP: What will it look like on the user end, and when will the service be available?
Stephan: We’ve termed our solution as a platform as a service. From the perspective of a lab director, if they want to unlock the cost savings and performance enhancements that come with next-generation sequencing, they don’t have to worry about doing any of the build out, with respect to sequencing or analytics. They can simply connect with us, we will turn on our platform as a service, and they can in a turnkey way, start doing next-generation sequencing and diagnostics. They can send samples or sequenced data to our central location, and we’ll do the analytics and send back reports without breaking CLIA chain. We’ll also have enterprise users. For example, at the Mayo Medical labs, we are installing our platform locally.
       We are taking preorders form labs right now. It will be available in the middle of the second quarter. That’s when we’ll have the platform fully available and launched.

Kurt Woock is associate editor of CLP.