Skip to main content
Premium Trial:

Request an Annual Quote

Genialis to Develop Biomarker-Based Cancer Diagnostic Models


NEW YORK – Central among American-Slovenian bioinformatics firm Genialis' plans for the use of its recent $13 million Series A funding round is to develop its own biomarker-based cancer diagnostic models without the assistance of outside partners.

"Those are going to become key products for us," said Cofounder and CEO Rafael Rosengarten. "[We'll be] taking biomarker models that solve problems for new and hot therapeutic areas and bringing those to market."

To date, the firm has used its technology to analyze clinical trial data for biopharmaceutical companies in order to inform future trial designs and to support diagnostic companies' commercialization of biomarker assays. This strategic shift would put Genialis in direct competition with Big Pharma, among others.

The company, which has twin headquarters in Boston and Ljubljana, Slovenia, has been developing cancer patient classifiers, based on machine learning and high-throughput omics data, to help predict response to targeted therapies.

Rosengarten described Genialis as a "computational precision medicine company." The firm's core ResponderID technology platform identifies and validates predictive biomarkers to support development of both diagnostics and therapeutics.

"We're really bridging the gap between these two rather disparate sectors of the biotechnology market," he said.

ResponderID is essentially a collection of predictive algorithms. One supports calling of tumor mutational burden from RNA sequencing data, while others call standard-of-care biomarkers such as microsatellite instability for certain cancers.

"The big vision is that through ResponderID, Genialis will be a platform technology company that links drug development diagnostics in such a way that every patient either gets the drug that is going to work on them best today, or from that patient we learn what drugs we need to develop to help them and others like them," Rosengarten added. "From a single tumor sample or from a single blood draw, a single sequencing run or sequencing assay, we can learn what we need to do to positively impact that patient's life."

Much of the success of artificial intelligence for diagnostic purposes so far has been with imaging analysis from firms including Owkin and PaigeAI. Rosengarten wants to be at the forefront of AI-based diagnostic development from sequencing data.

Competition includes what Rosengarten calls "do-it-yourselfers" — pharma companies with their own biomarker discovery and modeling operations. Other competitors include AI platform companies such as BC Platforms, Syapse, M2Gen, and Zephyr AI.

In some ways, Genialis is also going up against genomic sequencing labs that offer data analytics on top of their services, though Rosengarten sees companies like Caris Life Sciences, Ambry Genetics, Guardant Health, Tempus, and Roche's Foundation Medicine as "natural partners and collaborators" and said he expects to announce new partnerships later this year.

Rosengarten is optimistic that there is room for all kinds of companies in this space. "There are so many diseases to cover, problems to solve, input technologies to master," he said.

Genialis' key partner to date has been OncXerna Therapeutics, a Waltham, Massachusetts-based biopharma firm formerly known as Oncologie. Since 2020, the companies have been applying machine learning and neural networks to modeling gene expression signatures from RNA-seq data to evaluate and predict biomarkers for gastric cancer.

Rosengarten said that its work with OncXerna has allowed Genialis to validate its work in three areas: the technology itself, regulatory compliance with regard to bringing bulk RNA-seq to the point of care, and the business model that Genialis is pivoting to. In terms of regulation, Rosengarten said that Genialis has helped OncXerna gain US Food and Drug Administration approval for the use of AI as part of an upcoming clinical trial.

Genialis is a founding member of the Alliance for AI in Healthcare, and Rosengarten is a board member of that organization. Participation in the alliance allows a small firm like Genialis to keep up with regulatory developments and changes through dialogue with the FDA and the European Medicines Agency, he said.

OncXerna owns the 100-gene Xerna TME pan-cancer panel that it built with Genialis and has licensed that assay to major diagnostic firms including Qiagen and Exact Sciences, but the work opened Genialis to the possibility of developing its own diagnostic models and licensing them out.

"The technological validation, the regulatory validation, and the potential that there's a real market where drug companies and diagnostics companies want algorithms like this made us realize that we could do more if we had the funds to build these, not just for one customer but for ourselves," Rosengarten said.

Some of the Series A proceeds will go toward scaling up the commercial and operational side in the US, while other parts will support internal R&D activities.

In its announcement, Genialis said that the Series A totaled "more than $13 million" but did not get more specific. A Jan. 31 filing with the US Securities and Exchange Commission indicated that the firm had sold $11.4 million as of that date of the nearly $13.9 million in securities it was offering.

Rosengarten said that the company started raising money early last year and closed the round "a little while ago." Genialis waited to announce the Series A while it was finishing the validations of recent work, he explained.

While Genialis does have ongoing collaborations in neurodegenerative disease and immunology, its core internal R&D focus, fueled by the new funding, is oncology. "When we reach the next inflection point, what we're going to see is that our platform could scale everywhere," Rosengarten predicted. "I think that's where we'll start to branch out."

Rosengarten said the company will be looking to raise "a monster Series B" in the next year or two.

He referred to the firm's strategy as a "people-first" approach, which means concentrating on human biology even if the data is "messy," particularly in a clinical context.

"You have to really go out of your way to build datasets that are reflective of the intend-to-treat population," he explained. "Making sure our models work on real people is the first part of this."

"People first" is also a core value on the operational side, as he wants Genialis to be seen as a great place to work. The company currently has close to 30 employees, a "tech-heavy" group, according to Rosengarten, that includes bioinformaticians, data scientists, life scientists, and operational staff. Most of the technical employees are based in Slovenia.

Genialis is a very different company than it was at the time of its $2.3 million seed round in 2017.

"When I took over as CEO in 2018, we really pivoted hard in the direction of therapeutically relevant R&D," Rosengarten said. "We started partnering much more directly with clinical-stage pharma companies and addressing their problems around developing their drugs for diverse patient populations."

Genialis is trying to expand into target identification as it seeks to address conditions with no effective therapies.

While the company claims that its technology can identify biomarkers from essentially any data type, it is currently focused on RNA-seq data.

"We really like RNA sequencing for the time being," Rosengarten said. "It tends to be more phenotypic than DNA. We think it's better at representing those underlying [biological processes]. We think it tells a more comprehensive story about the difference between individual patients."

While RNA-seq has become commoditized from a laboratory perspective, Rosengarten considers it on the "bleeding edge" supporting regulatory submissions for diagnostic tests and devices.

Clinical data comes from public sources, pharma partners, and hospital partners that include Moffitt Cancer Center in Florida. Genialis Cofounder and Chief Technology Officer Miha Stajdohar said that the firm has patented technology to normalize data from disparate sources in order to address bias. Other patents cover preprocessing to prepare raw sequencing data for machine learning, he added.

ResponderID grew out of an earlier piece of software called Genialis Expressions, which still forms the core of ResponderID, acting as a data aggregator.

"As new datasets are onboarded to Expressions, their value increases exponentially," Stajdohar explained. A new collection of a few hundred samples might not be all that valuable on its own but add more value when combined with another 10,000 samples already on the platform.

"We can use other data to train the model and then validate the model on these few hundred clinical samples that we get from our customers, which we couldn't if you had started with those samples," Stajdohar said.