Nucleai is a company that uses AI and spatial biology to improve precision medicine and diagnostics by analysing tissue biopsies to predict patient responses to treatments.
Avi, tell us briefly about yourself and how Nucleai evolved from supporting oncology drug discovery to focusing on diagnostics within clinical trial settings?
I like to describe myself as a cartographer - for 20 years I mapped the world from satellite images. In 2018, I founded Nucleai with a similar mindset: instead of identifying objects from space, why not identify different cells under a microscope, including cancer cells and other components within tissue? That was the foundation of the company. From the outset, our goal has been to predict the right treatment for each patient.
Pharma companies spend years developing drugs, and one of their biggest challenges is determining which patients are most likely to respond. Many technologies attempt to do this, but they often ignore spatial biology—the exact positioning of cells within tissue. That’s what we bring to the table. By incorporating spatial context, we’ve raised prediction accuracy from around 30% to as high as 70–80%. We began by supporting pharma and clinical trials, then moved into active trial enrolment, identifying the right patients for studies. Ultimately, our focus is on developing clinical diagnostics that help physicians select the best treatment for each patient.
Your approach implies the increasing personalization of cancer treatment. Would you say oncology is definitively moving in that direction?
For sure. Like many areas of life, healthcare is becoming hyper-personalized. A treatment that works for you may not work for me, and biology is extremely complex, making these differences hard to predict. We need new ways to understand biology that account for its complexity, including spatial context.
We analyse not just genomic mutations but also where cells are located, their distances, neighbourhoods, and interactions within tissue. These spatial signals significantly enhance predictive accuracy and deepen our understanding of how a patient is likely to respond to different therapies.
So you are able to know in advance which drug will work for a specific patient based on their tissue mapping?
That is precisely what our platform is designed to do. Every cancer patient undergoes a biopsy, typically examined by a pathologist under a microscope. We take that same biopsy and apply AI and data from vast historical datasets to analyse it in far greater depth and accuracy. By comparing a patient’s tissue signature to those of past responders, we can predict who is more likely to respond to drug A versus drugs B or C.
This ability to determine treatment suitability a priori is the direction precision medicine is heading. Our platform makes such predictions possible at scale and with a level of accuracy that traditional pathology alone cannot achieve.
Could you share concrete examples from clinical trials where your technology has played a decisive role?
One example involves antibody-drug conjugates (ADCs), which function like highly targeted chemotherapy. We analysed cancer cells expressing specific proteins and discovered that the ratio between protein expression on the cell membrane and within the cytoplasm was critical—but invisible to the human eye. Using our platform, the pharma partner could predict with high accuracy whether a patient would respond to the ADC. That capability did not exist without AI.
Another case involves so-called “old school” bispecific drugs, which connect T-cells to cancer cells expressing two specific proteins. The challenge for a pathologist is determining whether both proteins are expressed on the same cell across millions of cells. We became the first company to be included in patient-enrolment criteria for a Phase 1 trial, using our platform to determine whether patients qualified for the study. Both examples show how we’re embedded directly into clinical drug development.
Do you expect your technology to apply to a wider range of therapies, such as cell therapies?
Yes. Our technology is modality-agnostic, which means it can support many types of therapies, including cell therapies, because it provides spatial context and shows essentially how different cells are arranged and interact inside a tissue. Anywhere doctors need to understand what’s happening within the tissue to guide treatment, our spatial AI can provide those insights.
While oncology is where these tools are used most today, the same approach can support other conditions too, such as inflammatory bowel diseases or certain skin disorders. If a disease relies on tissue analysis to make better decisions, our platform can add value. Expanding into those areas is part of our long-term plan.
When do you expect this technology to move from clinical trials into routine clinical practice?
Within the next two to five years our technology will transition from pharma R&D into clinical environments. Physicians will be able to use it for decision support, helping determine the best treatment options and better assessing a patient’s overall condition. That is our North Star: enabling real physicians to help real patients get healthier with more accurate and personalised treatment selection.
Personalized medicine can be expensive. How do you address concerns around cost and accessibility?
AI-powered diagnostics present a major opportunity to reduce costs and expand access. Traditional diagnostics involve labs, reagents, and extensive manual processes, which can be expensive. When diagnostics are driven by algorithms and computational tools, they become far easier to distribute widely, including in regions with limited laboratory infrastructure. For this reason, personalised diagnostics may ultimately become more accessible than many expect.
What is your relationship with U.S.-based companies, and do you foresee new partnerships emerging soon?
Partnerships are part of our DNA. We work with more than 20 pharma companies, most of them U.S.-based, including Gilead, Amgen, and others. We collaborate with them across the entire drug-development spectrum, from early mechanism-of-action studies to clinical trials. We also work with life-science tools companies—such as our recently published collaboration with Bio-Techne in Minnesota—and with academic institutions developing precision-medicine approaches, such as the University of Glasgow.
We intend to continue deepening and expanding these partnerships, as advancing precision medicine requires collaboration across the entire ecosystem.
You’ve spoken about what you call “the challenge of validation.” Could you explain what you mean by that and why it is so pertinent to life sciences?
When I talk about the challenge of validation, I mean that in many fields it’s hard to solve a problem but easy to validate the result. Sudoku is a good example—it takes effort to find the solution, but checking whether it’s correct is simple. When validation is easy, innovation accelerates because AI models can be trained and improved quickly. In life sciences, the opposite is true. Biology is incredibly complex, and even when we believe we understand it, validation requires clinical trials. You need to recruit patients, wait to see whether they respond, and that process takes a very long time.
A patient isn’t a computer—you can’t validate results instantly.
Meanwhile, technology moves at the speed of light. That’s the real bottleneck: embedding technology into the healthcare system takes years. If we could find ways to expedite validation using models, innovation in life sciences—drug development, precision medicine, digital health—could advance much faster. This point reinforces how our solutions can help address this ongoing challenge: speeding up interpretation, improving prediction, and reducing the inefficiencies that slow down drug development and clinical validation.
Looking ahead two or three years, where do you expect to see Nucleai?
Everything we do centres on helping patients. Everyone knows someone affected by cancer, and we want to contribute meaningfully to that fight. In the next two to three years, our goal is to help physicians and patients identify the right treatment more effectively so patients can achieve better outcomes.
All of our commercial milestones and company progress flow from that mission. Helping patients receive the right treatment is the foundation for everything we are building.
