Launching Phase I: 2026

AI-Native Immunotherapy

OpenIO is a federated framework integrating high-dimensional immunogenomics with generative AI. From descriptive immunology to predictive design.

Open ImmunoBank
Open Foundation Models
Open Protocols (SOPs)

 

Cancer immunotherapy has reshaped oncology, yet its efficacy is limited by patient heterogeneity and the complex dynamics of the tumor microenvironment (TME). Current reductionist paradigms fail to capture the non-linear context dependence.

OpenIO proposes a paradigm shift: from empirical screening to rational engineering.

  • Tokenization of Biology Treating cell states as tokens and tissues as context windows to leverage Transformer architectures.
  • Immune Scaling Laws Hypothesizing that model performance correlates exponentially with the diversity of cellular tokens.
  • Closed-Loop Validation Lab-in-the-Loop agents that formulate hypotheses and drive robotic experimentation.

The "Second Me"

The integration of data and models culminates in the Digital Immune Twin. Before a patient is enrolled in a trial, their "Second Me" undergoes high-dimensional simulations.

In Silico Cohorts
Simulating thousands of patient variations to model uncertainty.
Counterfactual Reasoning
"What if we used a PD-1 inhibitor instead of chemotherapy?"
Adverse Event Prediction
Detecting edge cases and toxicity before clinical administration.
Technology Stack

 

We are building a suite of AI models trained on the vast datasets within the ImmunoAtlas to learn the "grammar" of immunity.

Immune Language Models

Trained on millions of receptor sequences to predict antigen specificity, trace clonotypes, and design novel antibodies and TCRs. Moving from reading the repertoire to writing it.

Antigen & Presentation

Reconstructing the antigen-presentation pipeline from somatic mutation to MHC binding. Predicting which neoantigens are presented to guide personalized cancer vaccines.

Microenvironment World

Using Multi-Agent Reinforcement Learning where each cell is an agent. Simulating spatial organization and how barriers like fibrosis affect therapeutic efficacy.

 

Phase I: Foundation

2026

Establishment of ImmunoBank SOPs. Launch of first federated learning nodes. Open source release of immune foundation models.

Phase II: Generative Breakthrough

2027

Release of verified library of AI-designed biologics. First-in-Human study of fully AI-designed biologic in hepatocellular carcinoma.

Phase III: Autonomous Immunotherapy

2028

Self-evolving labs with closed-loop robotic experiments. "Second Me" simulation reports become clinical standard.

 

Fudan University
Stanford University
Neolife AI
Shanghai Jiao Tong University
MIT
ByteDance Seed
Shanghai Artificial Intelligence Laboratory
East China Normal University
Tongji University
Princeton University
University of Toronto
Zhejiang University
Westlake University

Join the OpenIO Consortium Now!

We are building the future of immune-oncology together. Integrate your lab into our federated network.

Clinical Partners

Contribute patient cohorts to ImmunoBank. Validate AI models in prospective clinical trials.

Join!

AI & Tech Labs

Co-develop multimodal foundation models. Build the "Second Me" digital twin infrastructure.

Join!