Scailyte is a spin-off from the Swiss Federal Institute of Technology in Zurich.
Making a diagnosis can be a tricky endeavour. For instance, diagnosing skin T-cell lymphoma — a rare form of cancer for which early detection can greatly boost survival — takes on average six years, because the skin rashes associated with it are difficult to interpret.
More common conditions can also be difficult to diagnose quickly. Endometriosis affects around 10% of women of reproductive age, but its painful symptoms are often vague and the condition is definitively identified only by an invasive laparoscopy. The average time to diagnosis is nine years.
Diseases with such slow diagnoses are prime targets for Scailyte (pronounced ‘skylight’) in Sursee, Switzerland. Chief executive Peter Nestorov describes the firm, a spin-off from the Swiss Federal Institute of Technology in Zurich (ETH Zurich), as a “biomarker factory”.
Scailyte pinpoints the cells that identify certain conditions by tapping into advances in two fields of research that are just starting to overlap. One is single-cell biological analysis, a field built by gathering data on millions of individual cells through methods such as RNA sequencing (for genome expression) or mass cytometry (for protein populations). The second is a form of machine learning known as neural networks, which can find needles in the high-dimensional haystacks of single-cell data.
Scailyte is built on CellCnn, a neural-network algorithm written by computational biologist Manfred Claassen, then at ETH Zurich, to detect biomarkers in single-cell data1 (see ‘Artificial diagnosis’). Promising results in 2017 suggested that the algorithm could find biomarkers that identify rare cell types with a role in various diseases.
Claassen, who is now at the University of Tübingen in Germany, likens the task to finding one piece of bad fruit in an incredibly tiny fruit salad. Conventional analysis blends the fruit and then tries to nail the culprit. CellCnn instead examines each piece of fruit in great detail. “CellCnn is much more sensitive than conventional approaches in directly taking the single-cell data and associating it with patient information,” he says.
CellCnn uses a type of neural-network architecture that has proved highly successful for image analysis, says Martin Hemberg, a specialist in computational genomics at the Wellcome Sanger Institute in Hinxton, UK. The 2017 study demonstrates that the approach also works well for single-cell data — in this case, data collected by mass cytometry. “CellCnn is highly sensitive and able to identify informative markers even for very small subpopulations of cells, such as for leukaemia samples,” Hemberg says.
In a study in 2019, CellCnn picked out a small subset of immune cells that are present at high levels in the blood of people with multiple sclerosis, but that declined substantially after effective therapy2. “That was quite remarkable, because the disease is not happening in the blood,” Claassen says.
Claassen formed Scailyte in 2018 with Nestorov, who had trained in biochemistry and genetics and at one point considered an offer to become a postdoc in Claassen’s lab. Instead, Nestorov took a job selling single-cell analysis equipment, made by the US firm Fluidigm, to academic and pharmaceutical research groups in Switzerland, but he stayed in touch with Claassen.
Finding a focus
The original aim of the start-up was to commercialize CellCnn as a tool for drug and diagnostic research. But as the pair enhanced and expanded its software resources, the company joined up with clinical partners to generate single-cell data from people with certain hard-to-detect diseases. “We knew as a start-up we needed to focus,” says Nestorov. “In 2019, we were successful in discovering a very sensitive biomarker for skin T-cell lymphoma, and we decided to focus on biomarker discovery in multiple disease indications.”
Scailyte is now creating a prototype clinical diagnostic application for lymphoma. This needs to be validated in large group of people to gather evidence for regulatory approval. Once approval is granted, the company expects to license the technology to an established supplier of diagnostic tests.
The spin-off has raised more than US$8 million, and plans to be pursuing biomarkers for ten conditions by the end of 2021, including skin T-cell lymphoma and endometriosis. Along the way, Scailyte is protecting its intellectual property by applying for patents on the biomarkers as diagnostic tools.
Scailyte will compete with companies that specialize in diagnosing specific diseases, and will increasingly cross paths with other firms that specialize in single-cell diagnostic work across multiple conditions. “The single-cell field is still an emerging technology, with new data modalities coming every three or four months,” Nestorov says. “That’s an opportunity because we have new avenues for discovery, but a challenge because we need to develop new computational tools to understand the new types of data.”
From a technical perspective, the company will face strong competition in this fast-growing field from many academic and commercial players, says Carsten Krieg, an immunologist at the Medical University of South Carolina in Charleston. Krieg used CellCnn as a research tool in a study on melanoma biomarkers3. “The more people get easy access to bioinformatic tools, the more difficult it is to create a unique product,” he says.
Staying successful will require not only remaining at the cutting edge of analytic tools, but also establishing and maintaining strong relationships with research hospitals and diagnostic suppliers, Nestorov says. In its first two years, Scailyte has made quick progress on all those fronts.