By Kevin Davies
January 15, 2013 | Silicon Valley Biosystems (SV Bio), an in silico diagnostics company founded by Dietrich Stephan and backed by the financial muscle of Sequoia Capital, announced its official launch today. The launch is highlighted by a report of a clinical case study in which a lengthy and expensive diagnostic odyssey was resolved when exome sequencing revealed the cause of two siblings’ neurological disorder.
Stephan, the firm’s founder and chief executive, says SV Bio aims to provide “rapid, accurate, and turnkey clinical interpretation of comprehensive genomics data at the point of care to improve the health and outcomes of patients.” The company hopes to catalyze the emergence of genomic medicine “by providing clinicians with the tools they need to make faster, more precise and cost-saving decisions.”
SV Bio joins an increasingly crowded field of commercial and academic groups eager to provide software tools for managing and analyzing patient genome data, including Knome, InVitae, Omicia, Ingenuity Systems, and the GeneInsight-Illumina Network. The competition will only get tougher in the coming months as other pedigree start-ups such as Cypher Genomics and Personalis are expected to reveal details of their genome interpretation offerings.
“All those groups… are spectacular,” a conciliatory Stephan told Bio-IT World. “I’ve known or talked to all of them, and believe they’re all on the right track. I think there are enough sick people in the world for 3-4 companies [to flourish], just like there are 3-4 major EMR companies in the world.”
In an interview with Bio-IT World last summer, Stephan told Bio-IT World that he hoped SV Bio would finish the work that his previous company, Navigenics, which he co-founded with USC oncologist David Agus, helped launch as one of a vanguard of direct-to-consumer companies in 2008. The firm quickly dropped the idea of engaging with consumers directly, in the face of stiff competition from 23andMe, and was acquired by Life Technologies last year.
“SV Bio is building the requisite bridge from the genomics bench to the bedside,” says Stephan, assisting physicians with the accurate and actionable interpretation required for optimizing individual patient care. SV Bio says it is offering a turnkey diagnostics service that allows a patient’s genome to be analyzed at the point of care, from gene panel to exome to whole genome.
Stephan says the clinical reports take a matter of minutes to complete from sample sequence to delivery. But SV Bio is not a pure software company – it operates three sequencing facilities: in Mountain View, CA; the east coast of the U.S., and South Korea. “When we started, we didn’t want to do anything with sequencing,” he says. “But in starting to accept raw reads from various places, we noticed a huge variability in the sequences. Being the ultimate owner, ‘garbage in/garbage out’ wasn’t anything we could tolerate.”
SV Bio features the Illumina HiSeq 2500 instruments, although Stephan declines to reveal the precise number in service. The company also has Ion Torrent machines “in the pipeline” and is watching the Ion Proton carefully. “Our business goal is to take technology decisions out of the equation and grapple with those so our customer don't need to,” says Stephan. “We think this has been a major issue in the field because the capital outlay is so large, the learning curve so steep, and the pace of progress of the technology so great that groups have been hesitant to make the investments and jump-start the field.”
Stephan says SV Bio has a capacity of 300,000 exomes this year, a figure that is “scalable in real time based on demand.” He hopes the firm’s sequencing facilities are “a temporary situation” and disappear in a few years, once the sequencing quality reaches a reliable threshold and the field has moved beyond short reads and the variable quality of hybrid-capture exomes.
Over the past year, Stephan says his colleagues have built a compendium of clinical utility case studies in order to benchmark the firm’s sequencing and software operation. In one case, a family was referred to SV Bio in which the eldest child had originally been diagnosed with a classical autism spectrum disorder. Over time, however, the child started to develop other neurological symptoms that were not associated with autism.
“That triggered a hugely expensive diagnostic odyssey -- $300,000 over two years in terms of testing genes, expert referrals to Stanford, Mayo Clinic, etc., and at the end of the day -- [after[ genetic, biochemical testing, and imaging -- no one had an answer,” says Stephan.
By the time he was contacted by the child’s physician, a younger sibiling had developed similar symptoms. SV Bio sequenced the exomes of the two affected children, and quickly revealed that they were compound heterozygotes for mutations in the TPP1 gene, associated with a rare form of Batten disease, a class of lysosomal storage diseases.
“The sequencing took a couple of weeks; we basically unlocked that [case] informatically in a day,” says Stephan. The gene defect was confirmed in a biochemical assay.
The results have given the family some clinical options, including trials either ongoing or about to commence for enzyme replacement or gene therapy. “It’s a fatal disease by age 8, the children would have died and it would have remained a mystery,” says Stephan.
SV Bio has not yet received insurance reimbursement for exome sequencing, but for good reason -- Stephan says he hasn’t sought it yet. “We’ve done the legwork, meaning creating a compendium of these case studies around clinical utility and cost savings. At least under the old CPT codes, we have a CPT stack that will reimburse us.”
The long-term vision of SV Bio – meaning 10-20 or 30 years – centers around the ubiquity of cheap, accurate genome information. “It will be a commodity,” says Stephan. “Many if not most people will be running around with their genome archived in their electronic medical record. We set out to build an in silico diagnostics company that can be invoked at the point-of-care by the physician on top of the archive genome.”
So how does Stephan get from here to there? “It necessitates grappling with the assay and providing that assay, given the genome isn’t pervasive. So we’ve launched with the ability to go from the full genome, which we believe will be a boutique offering… to the exome to gene panels.”
What will distinguish SV Bio from a bevy of genome interpretation start-ups nearing commercial launch, including Cypher Genomics, Personalis, InVitae, and several others? Stephan speaks of “key differences in how we architected the platform to make the information useful for clinicians.”
Stephan says there are three key words when it comes to the architecture of SV Bio’s platform: simple, accurate, and comprehensive.
“Simple”: “We’ve not created a genome browser with a pretty GUI on it that allows an individual browse through the data and really go on an exploratory journey. Rather, we’ve buried all the technological innovation in a very simple-to-use framework. What’s the problem you’re facing as a clinician or diagnostician? [You] push the buttons, and here are the genes and variants that are implicated in that disorder. We take all the expert knowledge that often exists within the academic or research setting to understand this complex infrastructure and boil it down, let the machine do that work, so the clinician can deal with the information and act on it.”
“Accurate”: “Accurate is central to clinical work – it deals with the notion of sensitivity, specificity, negative predictive value and positive predictive value, the classic terms in the diagnostic space but largely absent from many research-grade interpretative tools on the market today.”
Stephan says that current short-read technology on platforms such as Illumina still pose significant problems in recognizing insertions and deletions that are more than a few bases long. “If you compare that to Sanger, and expect NGS to supplant capillary electrophoresis (CE) sequencing, it’s not going to happen until you solve the sensitivity problem… Everyone is coming out with gene panels for cancer and inherited disease based on 100-250 bp reads, but the aligners, and variant callers (dozens of them) have trouble achieving the same sensitivity as CE.”
“In a nutshell, we’ve painstakingly benchmarked all the aligners and variant callers, tuned them all, and created an amalgamated solution,” he says.
SV Bio is also addressing what Stephan calls the false-positive problem. “If you’re a clinician, sitting in front of a mystery case, likely recessive, and dealing with 400 variants in 400 genes, that could absorb hours if not days/weeks/months trying to understand it. The notion of false positive compression… is key to accuracy. We’ve solved that as well with a proprietary set of algorithms that we’ve benchmarked in hundreds of cases with known lesions.”
“Comprehensive”: “This is a huge deal,” says Stephan. He cites, as an example, colon cancer. A physician interested in testing a patient with a family suspected hereditary non-polyposis colon cancer (HNPCC) can find numerous providers of the appropriate gene panel. “But the number of genes differs in every case!” he points out. “Some test 5, 7, others 19 genes. Which genes are the real genes? If I send my patent home and they’re mutation negative because I picked the wrong panel and they get colon cancer, that’s a big deal.”
SV Bio uses subject matter experts “to collate out bioinformatically the consensus genes for the phenotype of interest in a way the physician doesn’t have to bother with all the legwork. The physician will feel comfortable that they’ve got as good a snapshot as the current field allows.”
Stephan praises his current collaborators for their broad expertise in statistical genetics (UCLA associate professor Eleazar Eskin); machine learning/artificial intelligence (Sivan Bercovici) and computational biology (VP R&D Eugene Fratkin).
Funding for SV Bio comes from Sequoia Capital, a leading Silicon Valley venture capital firm that previously supported Apple, Google, Oracle, and Cisco, not to mention several biotech companies including Navigenics.
As for the possibility of analyzing complex diseases, as Navigenics originally offered in its Health Compass direct-to-consumer portal, Stephan says it is “not off the table.”
“I religiously believe there’s value in those genetic risk factors,” particularly for diseases such as age-related macular degeneration, myocardial infarction, atrial fibrillation and others, he says. “The science is real. The information has huge value. But the bump we hit at Navigenics is that it’s not reimbursed right now. It’s very difficult to drive clinical adoption where the focal plane of the event is relatively far out.”