Multivision Diagnostics - winner of the health and pharma pitching competition at Y Science 2023

On November 30, Y Science hosted a Health and Pharma Pitching Competition for early-stage startups and pre-startups. Today we had the opportunity to interview and , respectively the scientific and commercial leads of the project that won the 5,000€ prize. Multivision Diagnostics is a Finnish digital pathology innovation from the University of Helsinki which aims to aid pathologist and oncologists in finding the right treatment for each cancer patient. Below you can find out more about the story of this innovation.

This project started a few years ago as a purely academic project in . We had multiplexed staining images and we wanted to perform a very specific set of analyses to efficiently screen large cohorts of cancer patients. However, no computational tool was available to help us answer our biological questions, so we decided to build our own tools. In March 2020, in the week when the covid lockdown started, I remember biking to our lab to pick up the hard disk containing our data. Not being able to visit the lab during the lockdown gave me the time to focus on this project and so I started drafting the first set of codes and pipelines for the analysis. That same year, we decided we needed a computational scientist to really help make the analysis more efficient and that’s when joined the team.

By 2021 we already had some very promising proof-of-concept results, so Sara suggested that we could file an invention disclosure and start exploring the commercialization potential of this project. After discussing with (HIS, the tech transfer office of the University of Helsinki), we were advised to apply for the funding offered by Business Finland. HIS provided initial training to put together the funding application and helped us perform market research to make sure that our idea would indeed solve a need. Unfortunately though, our first application didn’t succeed and one of the reasons was that our team didn’t have enough business expertise, so that summer we started looking for a business developer that would join the project.

During fall 2022, Mikael got on board as commercial lead. At the same time, we joined the program, which put us in contact with several international experts and mentors from this industry who provided us with valuable feedback. All of this really strengthened our application, so when we reapplied for R2B funding, Multivision Diagnostics was awarded a 2-year grant which will end in December 2024. Now our small team of five is also supported by a steering committee of two business coaches, as well as a clinical advisory board with five oncologists and pathologists from different Finnish hospitals.

Currently, most cancer diagnoses still rely on generic staining methods and manual examination of glass slides at a microscope by a pathologist. Novel AI-based imaging tools are slowly entering the market, but these still mostly rely on the same traditional stains, and the training models require manual image annotations from a pathologist, a process that is very time consuming and which can be prone to bias.

Multivision Diagnostics represents the next generation of digital pathology tools. We combine structural analysis of the tumor tissue with information on up to twenty different protein markers expressed in the biopsy (using a method called multiplexed imaging), which gives us highly detailed information on the biology of each patient’s tumor. By comparing a new patient’s “tumor fingerprint” to those previously analyzed, we can make predictions on the patient’s prognosis and make recommendations on which treatment may be most effective for their specific type of cancer. Our technology relies on “unsupervised machine learning”, a method of analyzing large datasets which does not require time-consuming annotations from pathologists, and which is not as prone to human-introduced bias.

By working with clinical collaborators and Finnish biobanks, we have access to well-curated patient sample sets, allowing us to build large reference datasets for our two reference cancers: head-and-neck and colorectal carcinoma. In our proof-of-concept studies, we saw that our “tumor fingerprint” can identify responders and non-responders to certain treatments better than currently used diagnostic methods. Being able to correctly predict these two classes of patients could mean giving less aggressive treatments with fewer side effects to patients who are likely to respond well, while suggesting different treatments to patients with aggressive tumors who are unlikely to respond to standard conservative treatments. We are not yet at the stage to use our model in the clinic, but our first proof-of-concept shows promising results.

All this makes us confident that we can indeed build a clinical diagnostics tool that can aid pathologists and oncologists in developing the right treatment strategy for every patient.

While basic research focused on understanding more of science and biology is extremely important, it has been great to work on a project with translational potential. I think being part of this R2B project really broadened my way of thinking about research and science from a very different point of view. In academia, we usually focus on the novelty of the data and producing a high-quality publication, whereas when working on an innovation you also need to think about who will use your product, what is unique about it compared to your competitors and how you can turn research results into something that can be used in the clinic.

When you are the one working on a research topic, you are also the team most suited to take it further and actually turn your findings into an innovation. Publishing a paper isn't the end goal anymore. The end goal is to take the research results and actually use them to help patients, and that’s a very big shift. I am also discovering that such shift comes with a whole new workload requiring a new skillset, but the challenge of translating my research into a clinical innovation is something that I find quite motivating.

Y Science was a very positive and well-organized experience. Finalists of the competitions were provided a lot of training on how to build a pitch for investors, and this was particularly good for those who were more used to preparing scientific presentations and didn’t yet have much experience talking about their innovation from a business perspective. Joining both Y Science and Slush brought us a lot of visibility: we spoke with several people who may be interested in collaborating, and we also met with many potential investors.

During the coming year, a new software developer will join the team to help us combine the algorithms we have developed so far into a usable end-to-end medical device for clinical use. We have already filed one patent on Multivision Diagnostics’ platform technology for the analysis of large patient cohorts, and we are planning to file a second one on the complete set of biomarkers that we found to have the greatest predictive potential for patients in the head-and-neck cancer cohort, and potentially also in the colorectal cancer cohort. Meanwhile we are also preparing journal articles to publish all our validation data on these two cancers.

Since we plan to spin out by the beginning of 2025, we want to start talking to investors and ensure that we have adequate investment for when we will be launching our startup and start working on the clinical validation of our current results. In addition, we are looking into the regulations related to this type of technology, to make sure we understand what is required to take this project to the clinic as quickly and efficiently as possible.

Cancer is one of the most pervasive diseases of our time, and no family is unaffected by it. At Multivision Diagnostics, our goal is to transform the personalized treatment of cancer, so that each patient can have the best chance of beating their disease.