Prof. Derya Unutmaz: GPT-5 Accurately Predicted Complex Lymphoma Experiment, May Cut Discovery Time from Years to Weeks

Renowned immunology expert Prof. Dr. Derya Unutmaz, working in the United States, has highlighted the transformative potential of OpenAI’s GPT-5 in biomedical research.

In a social media post, Unutmaz revealed that he used GPT-5 to predict the results of a month-long, highly complex experiment aimed at engineering cells to fight lymphoma. The AI’s predictions, he said, were “astonishingly accurate”.

The experiment involved designing human CD8⁺ memory T cells to carry anti-CD19 CAR (chimeric antigen receptor), with about 90% of the cells containing a Tet-inducible CD19 antigen structure that only activates upon the addition of doxycycline. On day 0, CAR T cells were combined with CD19⁺ lymphoma cells. On day 2, doxycycline was introduced to trigger CD19 expression, and cell viability, activation markers, and proliferation were measured on days 2, 7, and 14.

“Predicted the Entire Process with High Accuracy”
Unutmaz noted that GPT-5 not only forecasted the entire experimental process and results with high precision, but also suggested design improvements. These included adjusting CD19 density to control fratricide (T cells killing each other), optimizing tumor-to-effector (E:T) ratios, and preserving an antigen-negative fraction for long-term cell persistence.

From Years to Weeks
He emphasized that this capability paves the way for fully in silico (computer-based) experiments: “This has the potential to cut scientific discovery timelines from years to weeks. A new era is beginning—one that will fundamentally change how science is done, bringing discoveries to the speed of thought.”