By Allison Proffitt
July 7, 2016 | Variant classification is a huge challenge—and a huge business opportunity. Cathy Tie, founder and CEO of Toronto biotech startup Ranomics, has a vision for the solution, and she has her elevator speech down pat.
“We have a proprietary technology where we’re able to assess how harmful each genetic mutation is in common disease genes in the lab,” she told Diagnostics World when we first spoke in April. “We’re able to generate these mutations and test these mutations in model organisms and human cell lines to understand how harmful these mutations are. We’ve created a database with the mutations we’ve tested in the lab and we’re currently helping genetic testing companies understand these mutations and really reduce the number of VUSs in these genetic tests, therefore making genetic tests more accurate and conclusive than ever.”
Ranomics accomplishes this by functional studies, Tie says. “We generate every possible single missense mutation for a gene,” Tie explained, clarifying that Ranomics does not use CRISPR. “Our technology is actually PCR-based,” Tie said, “an enzymatic process.”
“Once we’ve generated these mutations in the lab on a very high throughput scale, we would put these mutations into animal models or human cell lines: model organisms that can show phenotype that would show whether the mutation that we generated is harmful or not harmful,” Tie explained.
It’s an elegant solution if it works, but functional studies are tricky, and like many startups, Ranomics hasn’t offered peer-reviewed evidence that it has worked out the kinks.
When pushed for more specifics, Tie clarified that the “model organisms” Ranomics uses are yeast for the BRCA1 and TP53 assays. “We are also looking at expanding to human cell line assays for the BRCA1 and BRCA2 genes, but that is currently in R&D and we don’t have any updates for releasing those datasets,” she said in late June.
The database itself is the product. Through a web app, clinical labs that perform genetic tests can query the database for $100 per variant report or buy in bulk: $1,500 for 30 queries; $2,500 for 60 queries. Customers are limited to 60 queries per year. Ranomics has launched Ranomics RING Domain, a database of variants of unknown significance in the RING domain of BRCA1. The RING Domain database has 2,000 variants (each of the 20 amino acids substituted for the 100 amino acids in the RING domain of BRCA1). Ranomics plans to add “other hereditary cancer genes,” Tie said in our first conversation. Within the next three months, Tie told me in late June, Ranomics will release datasets for the BRCT domain BRCA1 and the entire TP53 gene.
For such an expansive vision, it’s a lean team. The company headcount is seven: two wet lab biologists, one software engineer, and four “business development people.” Tie, as founder and CEO, counts among the business development team members. The Board of Advisors is impressive: Mat Falkowski, founder of Invitae; Brian Naughton, founding scientist of 23andMe; and Karl Martin, CTO of Canadian software company Nymi.
“Functional studies in clinical genetic testing are still quite novel,” Tie said. “There still needs to be a lot of education done in this space. We’re very excited to be able to provide functional studies and to break this barrier in genetic testing and to enable more variants to be classified.”
Question Marks
Not everyone considers the approach novel though.
Lea Starita is a researcher at the University of Washington and has been doing functional studies on the RING domain of BRCA1 for five years. Starita was first author on a paper titled “Massively Parallel Functional Analysis of BRCA1 RING Domain Variants” (DOI: 10.1534/genetics.115.175802), published in Genetics in June 2015.
Functional studies are “not as easy as they look,” Starita says. She’s working in yeast—a good starting point, she agrees—and has recently started working in human cell lines. The transition has “caused a lot of problems.”
When I asked her about Ranomics’ claim to be able to test mutations in human cell lines, Starita laughed. “If I were one of these testing companies, I wouldn’t believe them.”
Doug Fowler, who runs a high throughput protein science lab at the University of Washington and is one of the pioneers of massively parallel functional assays, was also hesitant. “I’m skeptical, given how hard these assays are to do,” he told Diagnostics World. “But I don’t want to discourage them in any way. Maybe they have something!” he says. “It’s not that the idea is unsound. On the contrary, it’s exactly what we think should be done! But it’s a surprising claim.”
Starita’s concern is for the reputation of the approach. “We’ve been fighting in the field of genetics to get medical geneticists to believe us [that functional assays are useful for medical variant classification], but if someone goes out and sells mystery data to a bunch of people, our credibility goes down.”
For Starita, the caveat in this case is Frederick (Fritz) Roth, a respected professor at the University of Toronto and a senior scientist at Mount Sinai. It was in Roth’s lab that Tie began her research, and Roth now sits on the scientific advisory board of Ranomics. “The only reason to believe it is Roth,” Starita said.
So I called Roth to discuss the work. He was happy to discuss it, but opened our conversation with a disclaimer: “Look, I’m on the SAB [scientific advisory board] for Ranomics, which means if they want my advice, I can give it to them. But I can’t speak for what they have done or haven’t done.”
“I know they’re working on a BRCA1 functional assay in a human cell line using CRISPR, but I haven’t seen the latest on that,” Roth said.
Roth first met Cathy Tie in September 2014, her freshman year at the University of Toronto. As a high school student, Tie had done some biology research and some programming, and she approached Roth about working in his lab much earlier than he usually considers students. But he was impressed with her drive. “I asked her what her long term career goals were,” Roth recounted, “and she said she wanted to be the CEO of a biotech company. I just thought that was adorable.”
Tie’s freshman year she was working in Roth’s Mount Sinai lab, and there she met Leo Wan, a Ph.D. candidate in the Department of Molecular Genetics also working at Mount Sinai. Together they came up with the idea for Ranomics. They incorporated the company in January 2015, and in March Tie and Wan won one of 15 spots to travel to San Francisco to participate in the IndieBio Accelerator, a 100-day program meant to develop biotech businesses and secure funding. They arrived with an idea and a business plan.
In April, Tie was granted a Thiel Fellowship—$100,000 to put school on hold for two years and work on her business. In mid-June, Ranomics graduated from IndieBio with funding and interested customers.
Tie had just turned 19 years old.
“It’s a pretty cool story,” Roth says. “An ambitious young woman wants to be a biotech CEO and she just does it. I think that’s super cool.”
The Veritas Voice
It is cool, but it doesn’t validate the science. I asked Roth about Ranomics’ lack of peer-reviewed evidence. Is it really ready for market?
“That’s something they would have to provide the data to justify. I haven’t seen that,” he said. “[But] I’m not sure that publication is the yardstick by which a company should be measured. They need to convince their business partners, and if their business partners are signing up, they shouldn’t be giving away all of their tech.”
And the business partners are signing up. Veritas Genetics formally announced its partnership with Ranomics at the end of May. BGI came on board publicly at the beginning of June.
Joseph Thakuria, founder and Chief Medical Officer of Veritas, said his company ran a pilot study testing the Ranomics assay before it signed on as a customer.
(Veritas would not classify its relationship with Ranomics as anything other than a normal customer relationship. “However, we hope in the future that we will continue to gain early access to upcoming pilot programs as they are rolled out,” a company representative said.)
“We sent them some rare variants that we wanted them to use their yeast growth assay on, and the results were consistent with what we know those variants to be,” Thakuria told Diagnostics World. “They were consistent with the known pathogenic and benign classification of the variants.”
When Veritas announced the partnership with its press release, Starita asked both companies via Twitter if Veritas was aware that BRCA1 RING domain functional data were already published. Ranomics responded that it delivers “relevant data sources including published data” to its customers, and that a subscription to Ranomics “also includes access to our proprietary upcoming full-length BRCA1 and p53 datasets.”
Thakuria missed the Twitter exchange, but told Diagnostics World that in the testing phase Ranomics returned results “primarily on BRCA1 variants,” Thakuria said, “beyond” just the RING domain.
Veritas is currently using Ranomics as a last line resource for variants that don’t have other available classifications, Thakuria said. “If there’s sufficient data for the variant, there really isn’t a need [to consult the Ranomics database.] But as we and other groups do expanded panel testing and whole genome sequencing, it inevitably produces rare variants that have limited data.”
In March, Veritas announced its $1,000 whole genome sequencing service, which includes a report on any genetic variants Veritas finds that could affect customers’ risks of developing serious diseases, or have other implications for their lifestyle and health. There will be a lot of variants of unknown significance.
“We’re excited about the functional assay being used on other genes that have less data and clinical testing history, such as the numerous non-BRCA1 and 2 genes found on expanded hereditary breast and ovarian cancer panels,” Thakuria said. “The Ranomics data—in addition to other data like protein predictions, in silico tools—could potentially be important tools in classifying these data.”
The Open Data Argument
Of course Starita and others believe that if the classifications are right—if Ranomics does have a secret sauce for scaling up these kinds of experiments accurately—the results belong in the public domain.
“[Variant classification] is definitely an important problem,” said Jay Shendure, a researcher in the department of genome science at the University of Washington, via email. All three researchers are co-authors on the Genetics paper. “But my hope is that this kind of experimental data/models/interpretation all end up in the public domain, so that we don't end up with a situation such as currently exists with Myriad.”
Tie, for her part, insists that Ranomics is quick to give credit where public data is used, but that the Ranomics datasets contain proprietary data that aren’t available anywhere else.
“We recognize that [other groups] have done some really amazing research in this space and were the first to publish some results in the BRCA1 RING domain, and for a report that contained that data we reference them and give them credit for it. Our customers are very well aware of that,” she said. “A subscription to our databases includes not only the published data… but also proprietary data that we’ve generated in our lab that are not available anywhere else.”
Tie says that Ranomics’ work and that of academic labs should not really overlap.
“Most academic labs are more interested in discovery work,” Tie said. “What we do as a company is redo those concepts over and over again for every possible variant. That is not interesting to most professors.”
However, she does see areas for collaboration. “We’re very open to collaborating with clinical laboratories to publish some of the results that we have to validate our assay.”