By Allison Proffitt
August 7, 2013
| Real Time Genomics chose cows for their first proof of principle, but a lot has changed in the 18 months since I sat down in their Hamilton, New Zealand, offices and toured a dairy farmer’s cooperative down the road (see, Computer Scientists' Solution to a Biologist's Problem
). Today RTG’s extremely fast genomics analytics platform is proving itself faster, cheaper and more efficient than the competition for tackling Mendelian genetics.
The company now calls San Francisco home, and is headlined by a completely new leadership team and commercial focus. The founding CEO of Real Time Genomics, Graham Gaylard, left the company to attend to family matters and the company began looking for an “industry insider” to steer it into commercial viability, said current CEO Steve Lombardi. With an established biotech pedigree, having served as president and CEO at Helicos BioSciences, senior VP at Affymetrix, and VP of genetic analysis at Applied Biosystems, Lombardi was ready when Real Time Genomics approached him.
“I saw your article on RTG right when they were contacting me,” Lombardi told me. “Your article [aligned] very much with my thinking of them, which is that these are a bunch of really, really smart computer scientists, and with a little bit of pointing them in the right direction, their work could be really interesting.”
Real Time Genomics first vetted their technology with a local bovine genetics project, but the opportunities were much broader, Lombardi said. “I thought, boy, if this thing is good enough, we could apply it to the Mendelian genetics world. We could really have something!”
Lombardi took the reins in April 2012 (see, Lombardi Lays Out His Vision for Real Time Genomics
) and started building his core team. He asked Francisco De La Vega, an Assistant Professor of genetics at Stanford University and previous colleague of Lombardi’s, to look at the platform from a computational biology perspective.
“I remember [Francisco] said, ‘Steve, this thing looks really good, but there’s no validation. We don’t know if it’s going to work.’ And I said, Francisco! That’s why we need you!”
De La Vega signed on as Vice President, Genome Science in August. Pam Morley, a former sales exec at Applied Biosystems and Fluidigm, became head of sales in June. Jason Blue-Smith, previously a senior product manager responsible for BaseSpace at Illumina, joined the team in September as Head of Product.
The New Zealand computer scientists now work as part of a wholly-owned subsidiary of RTG. “The team in New Zealand has been absolutely fantastic,” Lombardi says. “These guys are real seasoned applied mathematicians and professional computer scientists. They share that wonderful sort of collective ego and sense of urgency that are needed to win.”
In the US, Lombardi and his new team have turned their sales, product, and research resources to human applications. RTG has two product lines; the first released to market is a shotgun sequencing-based metagenomics platform, primarily used by customers to “estimate species frequency composition and protein function, to understand the biology at play in a given sample,” Blue-Smith says. “The metagenomics was pretty much done before I got here,” Lombardi adds. “We have continued to work with a small number of customers doing some amazing things, but we just don’t think the market is ready yet for RTG to make a bigger investment in growing that application right now.”
Instead the company’s area of focus is the genome analysis platform—“RTG Variant” for now—which includes the Family Caller, Population Caller, Singleton Caller, and other potential products. RTG Variant technology delivers variants from raw sequence data or BAM files from large, complex pedigrees, or family trios for highly penetrant single-gene diseases to more complex, adult-onset diseases.
“A unique aspect of the technology is that the data from all available individuals can be jointly analyzed to not simply improve accuracy, but to detect types of variants that you can’t get by looking at a single individual,” Blue-Smith said. “But as importantly, this simultaneous joint analysis allows RTG to deliver comparable results when less sequence coverage is available. The implication this has on reducing sequencing costs for family and population studies is considerable.”
The first human validation performed with RTG using Illumina’s Platinum Genomes dataset revealed some impressive numbers. The analysis pipeline proved to be about 7 times faster than a comparable BWA mapping,” Lombardi says. “But where the rubber meets the road is in RTG’s ability to quickly determine variants. We’re 65 times faster… and we’ve got patented technology that allows us to do innovative things in the context of identifying actionable variants for different types of human disease.”
The newest validation is especially exciting for the field of early childhood disease, Lombardi says.
“Our technology can not only reduce the cost of analysis, but it also can reduce the actual cost of sequencing. The novelty of [the Family Caller] is that you bring all three aligned and mapped genomes into a single caller. What everybody else does is do each person in the trio—the mother, father, and offspring—separately… Because we do all three simultaneously—you’re using the mother and father to get as accurate view as you can off the offspring—you can reduce the coverage—i.e. the amount of sequencing you do on the parents—by half and still get the same results.”
RTG Family allows the actual sequencing coverage—and reagent costs—to be cut by 1/3. The team believes that’s enough to tip the scales for physicians who are stalled by the cost of exome sequencing for their patients.
“When we talk to pediatric clinicians and researchers about the ability to add parents without having to triple the cost, their eyes light up in the hope they’ll be able to provide better treatment options for these sick children,” said Blue-Smith.
The speed and accuracy both turn on the underlying mathematics. The platform has two components, explains Francisco De La Vega: proprietary algorithms for searching and a Bayesian infrastructure. The searching algorithms drive “really fast” alignments, and the Bayesian infrastructure is key because it deals well with prior information. “In the case of a pedigree, that would be the information that we know about the relationships between individuals and the expectations, for example, of Mendelian segregation.”
The infrastructure can integrate the sequencing platform error rates as well, producing a probabilistic quality score. “At the end of the day, the game of variant identification is about producing the right score,” De La Vega says. “We are constantly refining the balance between finding the true positives and avoiding the false positives.”
The platform is extensible and multi-threaded, De La Vega says. “It allows us to do variant calling with single individuals, allows us to do variant calling in populations, allows us to do pedigree variant calling, and, in fact, allows us to do variant calling in related samples such as tumor/normal samples.” Blue-Smith adds that the company soon plans to introduce separate products to deliver on the unique needs of each of these three use cases.
On a commodity server—“maybe $4,000 or $5,000”—Lombardi says a fully-mapped 30x human genome would take GATK about 65 hours to do variant calling. The RTG Variant Platform takes under an hour.
RTG’s goal is to be where the data is, says Jason Blue-Smith; the platform is currently deployed on the cloud, on appliances, and on customer’s local infrastructure. “It can run on a single server, on a distributed grid, in a public cloud; the point being that whatever IT infrastructure a customer uses we have deployed our products on and are able to integrate seamlessly into their ecosystem. No special hardware needed.”
The pricing is similarly flexible, Blue-Smith says. RTG offers pay-as-you-play pricing that scales with volume, requiring no up-front cost.
Lombardi’s vision is ultimately for the platform to be cloud-hosted. “My belief is that it’s all going to be moving to the cloud. As these NIH budgets get squeezed and health care costs get squeezed, people are going to get out of buying proprietary hardware in labs and they’re going to utilize the cloud.”
“The idea is to sell this not as a piece of enterprise-wide software—not even really think of it as software—just think of it as a consumable,” Lombardi continues. “If you’ve got a hundred exomes to run, you come to us and buy a hundred exomes. If you’re got 500 whole genomes, we’ll sell you that. If you’ve got a long-term idea of what your business is, you can come to us and we’ll give you more of a subscription-type of business. So we’re trying to be flexible.”
Lombardi sees nearly endless commercial opportunities for the RTG platform, but also sees a need for specificity. “The key thing you’ve got to do with a company is find a market where your technology can really be a winner, and then focus on it,” he says.
Lombardi has set his sights clearly on analysis. “Analysis is the new consumable,” he says. “Our 100% focus right now is to be a platform that transforms FASTQs into VCFs. When you look at sort of the sequencing value chain that way, you’ve got the sequencing companies whose main core competency is producing FASTQs, then you’ve got a lot of companies who are building either interpretation engines or full service CLIA labs to do the whole thing. But there’re just a few people who are trying to be the best at analytics.”
Lombardi mentions BINA, CLC Bio, and Novalign as having competitive pieces of the puzzle, but reiterates the same market position RTG claimed a year and a half ago.
“Our main competition, not surprisingly, is open source. Thirty-five years ago I was making DNA by hand in a lab; across the hall from us were people who were making enzymes by hand. Because no one would ever think of buying oligos or enzymes; we can do it best! Now you wouldn’t even think of it.
“What we’re coming now with is a value proposition, just like the DNA synthesis companies and the reagent companies and the sequencing companies and the microarray companies before, with a commercial product that is better,” said Lombardi. “We are as accurate or more accurate than the academic software and we’re much, much faster, and we’re easier to use. We’re bringing a professional approach to it.”
Lombardi’s confidence is supported by the product’s reception earlier this year. “We launched the product for the technologists at AGBT in February and got a great response. We went to the American College of Medical Genetics meeting in March and go an even better response.”
The buzz has borne out in partnerships as well. In a one month span from mid-April to mid-May, RTG announced partnerships or collaborations with Knome (to integrate the RTG Variant platform on the knoSYS 100 system); the J. Craig Venter Institute (a long-term study looking at the genetic changes that induced pluripotent stem cells may acquire during the process of differentiation); and Omicia (integrating the two platforms into a seamless workflow).
In May the company also announced a $5 million investment to further expand commercial operations. (It is currently backed by funding from Catamount Ventures, Lightspeed Venture Partners, and GeneValue Ltd.)
The funding, “gives us a nice stretch of runway,” Lombardi said. “We’ll continue to do what we’re doing, make further investments in our commercial franchise,”—the company is hiring in marketing, sales, and bioinformatics—“it’s only a matter of time until we breakthrough and really get going on this.”