From Data to Decisions: SkylineDx and the Art of Patient Stratification

Domenico Bellomo, PhD, Director of Bioinformatics at SkylineDx, fuses an engineering mindset with robust biological data to develop molecular diagnostics. Together with his colleagues at SkylineDx and academic and medical collaborators, Dr. Bellomo uses his expertise to bridge the gap between academically discovered gene expression signatures and commercially available diagnostic products that are ready for clinical adoption. As a result, SkylineDx is able to empower physicians with precise, molecular insights that enable personalized, risk-informed treatment decisions – ultimately improving patient outcomes. Patient stratification also helps drug developers identify which patients will benefit from novel therapies.

Engineering Precision Oncology

Although he completed his PhD work in electrical and control engineering, Dr. Bellomo now spends his days making sense out of large genomic and medical datasets. “Twenty years ago, there was a data deluge in biology, so there was a great need for people with a quantitative background – physicists, mathematicians, and engineers like me. Bioinformatics as a discipline was in its early stages, so I got lucky. I was already interested in biology, I liked reading popular science books, and then I had the chance to make it my daily work. Now, I use many of the same methods as I did as an electrical engineer but applied to biological data.”

Over the last decade, Dr. Bellomo and SkylineDx have grown together. “Twelve years ago, SkylineDx started as a spin-out from Erasmus Medical Center. I joined the company as a bioinformatician when there were only two of us. Now I lead the whole bioinformatics team,” he reminisces.

SkylineDx is built on a conviction shared by Oncode Accelerator: to advance patient care, we must better understand which patients will benefit from which therapies. Dr. Bellomo explains: “Historically, everybody wanted to develop a blockbuster drug that would work for every patient. These days, it’s all about precision medicine. We’ve realized that every patient is different, and not every patient will benefit from the same drug. That’s why it’s important to identify the subgroup of patients that will benefit the most from your new drug – to pay attention to patient stratification.”

Within Oncode Accelerator, the Artificial Intelligence (AI) platform is working on this challenge in collaboration with SkylineDx. Domenico explains SkylineDx’s approach: “With our work in molecular diagnostics, we try to answer three types of questions. First is the diagnostic question: do you have the disease or not? Second is the prognostic question. Once we know you have the disease, we want to know how well you will fare. And the third question is the holy grail: if we have treatment options A and B, which one should we give to you?”

The answer to that third question, “which treatment is right for you?”, is often complicated by factors such as patients’ genetic background, the potential to use different combinatory therapeutic strategies, and tricky cost/benefit analyses for specific drugs. Nonetheless, Dr. Bellomo’s dream is that one day, every new drug will come with a test that will tell you if you would benefit from taking it – allowing care providers to develop more personalized treatment plans and improving patients’ quality of life. This type of test is known as a companion diagnostic – and SkylineDx is actively working on developing them. SkylineDx also has a strong track record, with in-house developed diagnostic tests MMprofiler and Merlin Assay already commercially available.

When developing molecular diagnostics, SkylineDx proceeds with a mind for not only the underlying scientific intricacies but also the economic and regulatory challenges that come with product development. “Our starting point with each project is always one of those three questions – but there must also be an unmet clinical need. We also always perform an economic analysis, to assess competitors, eventual cost of the test, and if there is real societal need for it. We really bridge the gap between a discovery and the delivery of a test in the hands of physicians, to improve the patients’ outcomes.”

Hidden Challenges in Diagnostics Development 

To develop molecular diagnostics that address unmet needs in oncology as well as other indications, SkylineDx operates across both research and product development phases. “First, we need to find new signatures and biomarkers that can eventually become a product – that’s the research part. Then there’s the product development part. As a scientist, you might think that once you have identified a molecular signature it’s straightforward to develop a product. But if you want to make sure that the signature can be transformed into a test that becomes available for patients, you have to go through many steps within a highly regulated environment. The whole trajectory takes five to ten years.”

SkylineDx’s in-house knowledge of all these intricate steps from discovery through to regulatory acceptance is an asset to the Oncode Accelerator ecosystem, and can support drug developers executing Demonstrator Projects to develop more personalized cancer therapies. Dr. Bellomo elaborates, “We can start from scratch and be involved in the discovery phase as well as all the steps that come afterwards. In the discovery phase, we are agnostic as to the platform used to acquire molecular data – we can work with data from microarrays, qPCR, single cell sequencing… You name it. But we can also come in later, at the product development phase.”

When it comes to product development, according to Dr. Bellomo there are two crucial steps that are often not visible in the academic context: bridging the discovery platform technology – the technology used to identify a diagnostic molecular signature – to a test platform that will work in a clinical care setting, and the subsequent analytical and clinical validation of the resulting diagnostic test.

Regarding bridging, Dr. Bellomo explains: “The technology platform used for discovery can often not be implemented in the clinic. Therefore, you need to ‘bridge’ from the discovery platform to a technology platform that can be easily used in the average hospital. The test platform must work in the clinic on many different levels; it must be economical, time-efficient, and simple enough to pass regulatory review. For example, you may not expect it, but using single cell sequencing for diagnostic purposes can still be difficult from a regulatory perspective. Ideally you want to go to a simpler test platform, like qPCR or microarray.”

When it comes to analytical and clinical validation, in Dr. Bellomo’s experience, “This is something that people grossly underestimate.” In regards to analytical validation, he explains, “To deliver a test that is useful in the clinic, it must be robust and reliable, with results staying the same regardless of who runs the test, environmental factors stemming where the test is done – like temperature and humidity – and a host of other variables. You don’t necessarily think about this as a researcher.” 

Because SkylineDx has trod this path from discovery to development and beyond many times, they are able to start projects with the end in mind. It also helps that this expertise is spread across their team: “At SkylineDx, our whole team does a mix of both research and product development, which means we have both the in-depth scientific expertise and the big-picture understanding of all the steps that it takes to get to the clinic,” stated Dr. Bellomo. Dr. Bellomo also acknowledges the pivotal role of collaborations with academic researchers, clinicians, and other key stakeholders in driving forward projects and creating impact. “Without our collaborators, as relatively small company, SkylineDx would not be able to achieve all that we do.” 

SkylineDx x Oncode Accelerator: Harnessing the AI Explosion 

Within Oncode Accelerator, SkylineDx focuses on patient stratification – helping drug developers that are undertaking Demonstrator Projects to leverage AI and large clinical datasets to improve their understanding of which patients could benefit the most from the potential new therapy.  

“Our role isn’t in identifying new targets or therapies, but in personalizing the medicine,” explains Dr. Bellomo. “For example, it’s important but difficult to predict treatment response from survival data. Patients with poor survival might have benefited from their treatment, and their outcome could have been worse with a different treatment. Conversely, a patient with a good survival outcome could have had an equal or better response with another treatment. But many academic groups have developed computational algorithms to address this issue. Joske Ubels, who did a joint PhD between SkylineDx – supervised by our current COO Martin van Vliet – and the lab of Jeroen de Ridder, developed algorithms like STLsig and RAINFOREST to predict treatment response.” 

These algorithms make use of the so-called ‘virtual twin’ approach. “Ideally, to predict treatment response you would like to have clones of a patient, to test multiple drugs and see which one is the better match for the patient,” Dr. Bellomo explains. “We can’t make an actual clone, but we can create a virtual twin using existing data from populations of patients. It’s a simple idea: we can, for instance, measure gene expression of a given patient, and then look for individuals within clinical trials who had similar gene expression profiles and see how they responded to treatment A versus treatment B.” In this way, drug developers can make a more educated guess at the likely response of the patient.

Jeroen de Ridder, Professor at UMC Utrecht and Oncode Investigator, is now a work package leader within the Oncode Accelerator Artificial Intelligence (AI) Platform, and Dr. Bellomo has continued to work closely with him. Through the AI Platform, the scientists are again co-supervising a PhD student, Lisa de Groot, who is further exploring AI-based approaches, including foundation models, to develop gene expression signatures that can predict treatment benefit in cancer. “With the boom of AI, Lisa is now building on the work we started together with Joske – using newly available data and AI-driven methodologies to tackle the challenge of patient stratification.” Thanks to their existing expertise in molecular diagnostics and patient stratification alongside this enthusiasm for exploring new and cutting-edge methodologies, SkylineDx is a valuable partner within the Oncode Accelerator oncology ecosystem.

About SkylineDx 

SkylineDx is a biotechnology company focused on research and development of molecular diagnostics in oncology inflammatory, and infectious diseases. SkylineDx uses its expertise to bridge the gap between academically discovered gene expression signatures and commercially available diagnostic products with high clinical utility, assisting healthcare professionals in accurately determining the type or status of disease or predicting a patient’s response to treatment. Based on test results, healthcare professionals can tailor the treatment approach to the individual patient. SkylineDx is headquartered in Rotterdam, the Netherlands, complemented by a US base of operations and a CAP/CLIA certified laboratory in San Diego California, USA. To learn more about SkylineDx, please visit www.skylinedx.com.