In the rapidly evolving landscape of immunology and drug development, a new paradigm is emerging—one where artificial intelligence and deep immune profiling converge to decode human health and disease. Immunai is at the forefront with its ambitious “Grand Collaboration” initiative. Last summer, the company announced a call for proposals to partner with leading academic institutions on groundbreaking immune-centric studies. The goal was clear: to accelerate translational medicine by generating profound biological insights from rigorously collected clinical data.

Now, with four specific projects chosen, the work enters a critical phase. Drug and Device World caught up with Noam Solomon, co-founder and CEO of Immunai. He delves into the details of these collaborations, spanning autoimmune kidney disease, cell therapy toxicity, novel combination immunotherapies, and gastric cancer. He explains the strategic thinking behind selecting a diverse portfolio, how these projects feed into Immunai’s growing Annotated Immune Cell Atlas (AMICA), and his vision for a future where AI-driven insights enable truly personalized, safer, and more effective treatments. This conversation offers a rare glimpse into the operational and scientific strategy of a company aiming to map the immune system in its entirety.

This interview has been edited for clarity, consistency, and length.

Phalguni Deswal [PD]: Last time we spoke, you were collecting submissions for your Grand Collaboration. Now you’ve selected four distinct projects. Were you intentionally looking for studies covering different patient populations, or did it naturally unfold that way?

Noam Solomon: It’s a very good question. The criteria for selecting the projects were quite diverse. It related to the principal investigator—their track record and the research they proposed. Central was the relevance to the immune system; we wanted studies on immune-modulating agents. We also consciously wanted a diversified portfolio. It wouldn’t be interesting or a true presentation of our capabilities if all projects were the same drug in the same disease. We want the world to see what can be done across a variety of indications and therapeutics.

PD: These projects aim for lasting impact. Let’s start with the first: autoimmune kidney disease with Mount Sinai. The project is comparing immune signatures across conditions with underlying immune dysfunction. With the other three in oncology, what’s the intention? To test if regulatory pathways are common across autoimmunity?

Noam Solomon: We actually had more applicants in oncology, but you select from the best proposals. The aim was to explore both inflammation/immunology and oncology. There’s something super interesting about studying the immune system across different autoimmune or inflammatory conditions. In many indications, you target the same pathway. So, studying psoriasis, atopic dermatitis, and alopecia, or asthma and COPD, reveals common elements. We are trying to map and discover common and different immune signatures across immune dysfunctions to see which are specific to one indication and which are common to many.

PD: The second project on cell therapy toxicity prediction is particularly interesting. How do you expect this to inform and de-risk cell therapy development?

Noam Solomon: Cell therapy is one of the most innovative frontiers in precision medicine but is challenged by toxicities like cytokine release syndrome (CRS), ICANS, or HLH. These can be dangerous and complex. When developing an expensive new cell therapy, if you could predict the sources of toxicity early, or identify biomarkers for which patients are more likely to have a toxic event, you dramatically increase the chances of clinical success. It helps monitor patients and avoid terrible events. There’s a scarcity of clinical data in this area. This study aims to yield results that will be impactful for many future cell therapy projects.

Furthermore, if we can deconvolute different toxicities and predict specific types—not just “any toxicity”—that’s incredibly valuable. It’s work only a granular platform can do. Interestingly, CRS also occurs with drugs like T-cell engagers. Insights from one area can apply to another. That’s our vision: mapping the immune system across a diverse set of indications so that learnings from one drug can inform findings in another.

PD: The third project, on neoadjuvant immunotherapy in NSCLC, aims to build insights into mechanism of action, response prediction, and adverse event modeling. Is the long-term goal not just determining the best combo, but personalizing combination therapy for individual patients?

Noam Solomon: That’s a good question. There’s a difference in what “personalized” means. Autologous cell therapy is extremely personalized but expensive. Another version is personalizing the combination or dose for a patient using existing medicines. A good outcome is defining a specific treatment regimen based on combination, dose, etc., without inventing something new per patient.

Currently, clinical trials aren’t adaptive enough for this—they don’t even easily accommodate differences between men and women. There’s a call to action to allow more adaptable trials where AI tools can modify treatment based on a patient’s unique profile.

Right now, we’re answering questions at the cohort level: Which antibody combination is better? Which dose? It’s personalized at the group level. The same methodology can define a biomarker threshold: if you’re above it, you get this dose; below it, that dose. For example, a higher dose may increase efficacy but also toxicity. If we can predict a patient’s likelihood of toxicity versus lack of efficacy, we can personalize the dose. Five or ten years from now, I envision clinical trials allowing more precision medicine earlier, with every decision on combination and dose informed by patient-specific attributes. Our platform is uniquely fit for this because we can analyze a blood draw and recommend the right path.

PD: Moving to the fourth project on gastric cancer—you already have experience in colorectal cancer. Is this about expanding your GI cancer data platform and understanding resistance mechanisms?

Noam Solomon: Yes. Generally, the more FDA-approved drugs in an indication, the easier it is to build data foundations. We’re building a gastric cancer data foundation because it’s a major unmet need. It’s a prevalent cancer with limited therapeutic options.

You also asked about our IBD collaborations. Few companies try to map GI health end-to-end. They might study IBD or colorectal cancer in isolation. We’re asking if there are relationships between them. People may be unaware that some patients on oncology drugs develop autoimmune conditions like colitis as a side effect. Studying these areas in concert is very important.

Sometimes, to understand one thing deeply, you need to understand related systems. To study cancer, it’s useful to study patients before they get cancer or those with related immune dysfunctions. Very few companies take this holistic, agnostic approach to model different immune states. The first immunotherapy, anti-CTLA-4, was discovered through a genetic mutation linked to autoimmunity. That inspired Immunai’s mission: to map the immune system with AI. By bringing in more indications, therapeutics, and data on responses and adverse events, we can uncover insights that cross traditional disease boundaries.

PD: You’re working with premier institutions and scientists, generating rich data meant to serve needs beyond these specific projects. How are you thinking about data standardization and scalability?

Noam Solomon: Why work with the best PIs? This isn’t a machine where you just press a button. First, it’s about asking the right questions. Aligning with top researchers increases the chance of success. Second, it’s about the qualitative procurement of biological specimens. Good investigators are good physicians; they know how and when to take the right measurements. Garbage in, garbage out. Third, it’s the clinical annotation—accurately recording if a patient is responding, side effects, etc. High-quality samples and annotation are crucial for extracting meaningful signals.

These four studies give us confidence in the investigators, the research questions, and the sample quality. I’m optimistic we’ll find things, some specific to the project, others relevant across our entire database.

PD: You have reiterated these collaborations are for profound biological insight that feeds into asset creation. How do you ensure that happens?

Noam Solomon: It’s straightforward. All the data from these studies—hundreds, eventually thousands of samples—will be integrated into AMICA, our Annotated Immune Cell Atlas. The harmonization is done with the same AI and organizational scheme. The strength of the insights we generate for our partners is a direct result of this integration into AMICA, the database we’ve been building and growing for years. This unified approach is what allows us to generate productive, high-quality insights that advance the field.