Using Artificial Intelligence to decode the immune system for precision oncology and early cancer interception.

The immune system is an extremely sensitive warning sensor, constantly scanning for infections and cancer. Its responses often decide whether we recover or fall seriously ill. Our mission is to bridge the gap between century-old blood tests and modern single-cell technologies using artificial intelligence to transform how we understand, diagnose, and treat disease.

In oncology, we still rely on crude insights from traditional blood tests (like blood counts and CRP). We lack precise ways to use the immune system's rich information to guide treatment. However, new single-cell technologies now allow us to examine immune cells in unprecedented detail. Combined with artificial intelligence, they hold the potential to reveal how cancers arise, how to intercept them early, and how to tailor therapies to individual patients.

We address this gap by linking clinical trials, experimental systems, and AI:

• Interpretable AI: By leveraging prior knowledge and large data repositories, we build tools that generalize to thousands of patients.
• Patient-Derived Validation: We validate our insights into how the immune system eliminates or fosters tumors using patient-derived experimental systems.
• Multimodal Integration: We build frameworks to integrate these insights into existing clinical workflows, incorporating electronic health records (EHR), radiology, and pathology data.

We have established that clinical outcomes can be predicted directly from blood immune responses. We are leveraging these insights to build a new generation of tools: Artificial Intelligence Immunodiagnostics.

These precision tools can:

• Distinguish whether inflammation is driven by infection, autoimmunity, or cancer.
• Determine if a patient needs antibiotics, immunosuppressive treatment, or adjusted cancer therapy.
• Highlight new therapeutic targets to strengthen anti-tumor immunity and intercept malignant disease early.

We develop AI immunodiagnostics to predict patient outcomes from immune cell states in the blood. By combining single-cell transcriptomics with domain-aware machine learning, we identify conserved gene programs and link immune activity to therapy response. This moves us beyond simple blood counts to capture intra-cell type heterogeneity, turning the immune system into a ubiquitous biosensor for cancer, infection, and other health exposures.

How does human immunosurveillance shape the earliest progenitors of cancer? Moving beyond mouse models and invasive tumors, we study premalignant immune control in humans. Using Lynch syndrome as a model with defined antigens and paired blood-tissue samples, we integrate single-cell genomics, spatial transcriptomics, and patient-derived "avatar" mice. This reveals how immune evasion develops stepwise and identifies the optimal window for reversing it.

We use multimodal AI to understand how the immune system links diverse clinical conditions. Our models integrate radiology, pathology, dermatology, gene expression, and immune profiling data with electronic health records. This framework goes beyond simple prediction—it uncovers underlying immune mechanisms across diseases to enable true precision prevention and treatment.