GPT‑5 Pro Solves 3-Year Immunology Mystery

Immunologist Derya Unutmaz used GPT‑5 Pro to solve a three‑year‑old puzzle about how glucose affects T‑cell specialization, revealing that deoxyglucose interferes with IL‑2 production and enables Th17 differentiation.

By uploading experimental data into GPT‑5 Pro, the model suggested a mechanistic insight that was not evident from the lab data alone, and it correctly predicted the outcome of a subsequent experiment involving CD8+ T cells and lymphoma killing. The model’s ability to generate novel hypotheses and simulate experiments demonstrates a new level of AI collaboration in scientific research. Unutmaz highlighted that GPT‑5 Pro can accelerate research by reducing weeks or months of work, but also emphasized the need for responsible AI safeguards to prevent misuse. OpenAI’s Preparedness Framework outlines risk mitigation strategies for advanced AI capabilities that could be used for biological weapons. The study showcases how frontier AI models can function as research collaborators, providing insights that would otherwise require extensive domain expertise. Unutmaz’s experience illustrates the potential for AI to transform hypothesis generation, experiment design, and data interpretation in biology. The paper also notes that GPT‑5 Pro’s predictions were not based on publicly available data, underscoring the model’s internal knowledge base. This breakthrough demonstrates the practical impact of advanced LLMs on accelerating scientific discovery. The work raises important questions about governance, transparency, and the ethical use of AI in research.