Aignostics, a Germany-based spin-off from Charité – Universitätsmedizin Berlin which specialises in AI-powered pathology, has closed a €14 million Series A round. The round was led by Wellington Partners (German VC which has invested in another unique European health tech innovator Sidekick Health). They were also joined by existing investors Boehringer Ingelheim Venture Fund (BIVF), IBB Ventures’ VC Fonds Technologie, High-Tech Gründerfonds (HTGF), and by new investor Ascension.
Individual pathologists are exceptional at interpreting single samples. However, this highly focused skill is less effective when it comes to analysing large sets of samples, making the identification of characteristics and biomarkers that can inform research difficult and time-consuming, if not impossible.
Aignostics, which has now raised almost €20 million in funding, plans to further develop its AI models, and expand the company in Europe and into the US market. Another one working in this field is Deciphex which pocketed $11.5M recently to help pathologists work efficiently using AI.
Meanwhile, Viktor Matyas, Aignostics’ CEO and co-founder, joined TFN to share their story and plans.
A foundation in academic research
Matyas explained that the company had a long gestation period, beginning over ten years ago when Aignostics’ senior pathologist and co-founder, Professor Frederick Klauschen was working at the Charité – Universitätsmedizin Berlin. “He spent, perhaps, 60% to 80% of his day staring through a microscope,” Matyas said. “And he realised that perhaps that’s not the most efficient way to do it.”
It resulted in a collaboration with Professor Klaus-Robert Müller who was working on machine learning at the Technical University of Berlin which built the foundations of Aignostics. Matyas, and the fourth co-founder and CTO, Dr Max Albert, joined in 2018. Starting in the Charité – Universitätsmedizin Berlin accelerator programme, Aignostics was formally launched in 2020.
This background has given Aignostics its key advantages in the market, access to pathologies and close ties with the research community. Lead investor Wellington Partners’ Principal Dr. Johannes Fischer said, “what makes Aignostics unique is not just its technology, but also its comprehensive access to key data modalities and pathologists, who are closely involved in the development process at all times.” He continued, “this combination allows Aignostics to rapidly build bespoke pre-trained AI models that are highly impactful when used by its clients on their own research and clinical trial data.”
A different approach to AI
Medical AI has not always had good press, with famous examples including diagnoses based on metadata rather than clinical information. However, Aignostics have patented an “explainable AI” solution. This reverse engineers their models, highlighting what its decisions and predictions are founded on. They have also developed a model that does not rely on pathologist annotations, meaning learning takes place only on the samples provided.
Together, it promises a change in the way that pathology is approached. “The problem is that tissue samples, under the microscope, are incredibly difficult to analyse,” Matyas said. “You need to zoom in so much that you end up looking at maybe 1% of the actual tissue sample. And the pathologist can only look at it in a very qualitative way.” In a clinical trial that might have thousands of samples, it’s reliant on estimates and pathologists noticing correlations. This can mean that crucial insights are missed and, until now, that much of the diagnostic and therapeutic information could remain unseen.
Growing the market
The company is currently 55-strong, with a further group of around 20 pathologists who are working closely with them to help develop their models. Berlin-based, their current client list has a European bias. However, they plan to expand into the US market.
“We want to do more in the US and have hired a very experienced Chief Business Officer, based in Boston,” Matyas explained. They are also looking at moving beyond research projects. “We want to get our platform ready to be able to host prospective clinical trials. Today, everything we do happens on retrospective data. The plan is that we will take these models into clinical routine.”
They are currently working on developing their platform to enable them to take part in clinical trials, putting Aignostics in a strong position. “As a first step, we want to run these models in early-stage clinical trials,” says Matyas. It will make them a market leader, “we want to produce companion diagnostics and complementary diagnostics. Today, there’s not a single tissue-based, AI-based companion diagnostics out there.
Better treatment through AI
Matyas sees two types of companies working in the AI field, those that focus on the individual samples, essentially assisting or even replacing pathologists, and those like Aignostics, that focus on research and clinical trials. “We’re focusing on the pharma research side,” he says. “We think it’s more exciting. And you can really do more for the patients in the midterm.”
The potential is there for a significant change to the speed and accuracy of research. While how long it takes for Aignostics, and AI more generally, to appear in clinics will depend on regulatory approvals, Matyas is confident it will be sooner, rather than later. “I think AI is definitely here to stay for research and for clinical trials and in a few years down, you will not see a clinical trial that’s not, at least, being quality controlled by AI.”
