By Kevin Davies
January 28, 2013 | The genome interpretation field gets a little more interesting this week with the news that Cypher Genomics, a software start-up co-founded by a notable quartet of Scripps Institute physicians and scientists, is launching its beta access program. The San Diego company was launched by cardiologist Eric Topol, mathematical geneticist Nicholas Schork, bioinformatician Ali Torkamani, and neuroscientist Ashley Van Zeeland, the company’s CEO. Bio-IT World editor Kevin Davies recently spoke to Van Zeeland to learn about the company’s goals and core strengths.
Bio-IT World: Why is clinical genome interpretation such an appealing field for a new business?
Van Zeeland: It’s an enormous problem and any big problem is fun to work on! The other reason it is so critical is it’s a scale problem. You can’t functionally validate every VUS [variant of unknown significance] in an individual’s genome or across populations, so we have to come up with other tools to address that need.
From our perspective, our interest is trying to get the most accurate interpretation of those VUSs—aggregating results across information sources to come up with an accurate prediction. All of this is state of the field—but as more genomes are sequenced and we learn which variations are tolerated and not tolerated, then I think we’ll start gaining some traction. But for now, we have to do the best we can.
Are you focusing on just clinical implementation or do you see research applications too?We see this as a virtuous cycle of sorts. If we can move this field along and enable additional discoveries through research sequencing, then [move] into the clinic, that’s ultimately the goal. We are building a clinical interface for use by hospitals, but we also have the scale to support very large research studies. Because of [co-founder] Nik Schork’s involvement, we have the expertise to run these large scale studies.
There are many software companies looking at the genome interpretation space. What sets Cypher’s technology apart?That’s difficult to say—I get 2-3 emails a month from colleagues saying, ‘have you heard of this new company?’ It’s crazy! Most of them are so nascent it’s hard to judge at this point. We’ll have to let them come out and let the market decide.
There are many different aspects of genome interpretation and among existing players and new players, I can see different market fits. Certain groups may be really good at family-based interpretation. Cypher has a particular strength in VUSs and using systems biology to understand the genome. Other groups are focusing solely on big case-control studies. And there are big players focusing on cancer profiling.
There are so many ‘interpretation’ companies in the genome analysis space that confuses the issue further: Some folks are doing alignment and variant calling and that is called genome interpretation. Some are doing well in visualization and that is also called interpretation. Cypher is somewhere in the middle—from variants to what you visualize—but I’m interested in whether the market is going to look for strong players in each of those fields or if it will want one entity that will do the whole thing?
Going back to your question—what sets us apart—we have paid extreme attention to security in building the system. We specifically did not build on certain cloud platforms because of the security risk.
Also, I’d point to the speed with which we can do a complete annotation of a genome. We can go from variants to deeply annotated analysis in just about one hour at this point—across all 3 to 4 million variants in a genome.
You mentioned systems biology. How does that play a role in your offering? We use systems biology principles, looking across a whole-genome view, both for interpretation of idiopathic cases to identify likely disease-causing variants, as well as in large statistical studies to look at collections of rare variants that may fall across the genome in different categories—rather than just a priori go after different sets of genes or functionally damaging variants. So we try to link it all the way to cellular physiology.
A key issue in genome interpretation is the quality of clinical databases. Is Cypher working on any new proprietary clinical variant databases?We are working with partnerships in building out certain clinical databases, we recognize value there. Everyone sees, at the end of the day, what that’s getting at is sequencing more genomes. Part of what we’re doing, in collaboration with Scripps Health and Complete Genomics, is sequencing a thousand “healthy elderly” population—a database of individuals to mine the data and understand what variants are tolerated and what ones are not. So that’s just one example of how we’re moving in that direction.
No matter how rigorous your analysis is, how important is it that physicians can access and make sense of the final genome interpretation report?
I think visualization is key. Some companies have great visualization tools and it’s something we’re paying a lot of attention to. We’re also talking to potential partners about seeing what our strengths are and what their strengths might be, to leverage the best of both worlds. So we’re talking to other groups while building out our own tools, trying to make [the report] as undaunting to understand as possible. There’s a lot of information—at one level you want to communicate the gestalt of that genome, but also zero in on the important pieces of information within it. That balance is important, and you can get lost in the weeds.
You have an impressive group of co-founders. What do they each bring to the organization?I’m incredibly fortunate to have found myself in the middle of this team. Eric Topol brings the genomic medicine/physician’s perspective, real-world clinical experience and drive to change medicine.
Nicholas Schork brings expertise in population genetics and cancer genomics, specifically the nuances of looking across multiple genomes and the statistical expertise.
Ali Torkamani serves as our CSO right now. This is really his brainchild—the systems biology principles, cancer biology, comes from his expertise, as well as the bioinformatics and architecture, and our ability to speed this up. He’s really leading the technical development.
I’m a co-founder—I’m a neuroscientist who wanted to get into genomics. I used this tool in my own research, after banging my head against a wall doing my own bioinformatics. This tool made the information accessible. I saw the value in that—it let me interpret the genome rather than spend my time coding. I also have a business background, so I’ve been leading the commercialization effort.
What’s the plan for launching commercially in 2013?We’re actively building out a user interface for web access to our tools. We’ll roll out the first analysis module as part of a beta program soon—and there are multiple things you can do with the annotated data, including running clinical filters, sharing genomes across accounts, and combining individual genomes into larger analyses that will be introduced during program. We are announcing our early access program that will start in February at the Personalized Medicine World Conference, and we’ll be going to a number of conferences demonstrating and getting feedback, expecting a commercial launch in late spring.
Are you planning to license the software to users or provide a web interface while running on your own cloud?We’re tilting in favor of using our own systems. We knew there would be some hesitancy about this, so as I said earlier, we’ve paid close attention to security. We want people to be confident that their data are protected, so we have multiple levels of security around the data. But we’re also working with partners who want some piece of this software behind their firewall. So we can build a hybrid approach for those customers as well, where a lot of the data can remain local, but we leverage our high-performance compute system to do some of the heavy lifting.
Did you decide against using the Amazon Cloud?Yes—although many folks have gone with the Amazon platform, we’re working with another commercial provider to effectively build our own private cloud with them. Everything will be behind a firewall, all the threat detectors, independently connected to our own cloud. We appreciate the concerns about these data getting out. I’ve heard stories from other cloud providers where someone has spun up a cloud server and there’s still an image of another session there. For genomic information, that’s unacceptable! And it’s also unacceptable to have access to that information often go down for a while. So data replication across different data centers is very important to us.
In general, what are the main challenges in getting whole genome/exome interpretation widely accepted and reimbursed?I’m not at the point where I think it’s inevitable, but it’s worthwhile and there are some elephants in the room that need to be dealt with so that adoption is possible. You have to make the economic case—Howard Jacob [Medical College of Wisconsin] has done an amazing job in that regard… and I think the fear of ‘the incidentalome’ has to be addressed first. It’s hard to prove a negative—that is, that without genome sequencing the clinical course would have been different. When we have a better handle on the incidentalome and the economic benefits—instead of doing endless rounds of single gene tests—we can just look at the genome. I think it [widespread whole-genome sequencing] could happen—and it should happen, frankly.