GPT-5 solved a 3-year immunology mystery human researchers couldn't
openai's gpt-5 pro identified a protein interaction mechanism that baffled immunologist derya unutmaz for three years, signaling ai's evolution from pattern matching to scientific discovery.
jackson laboratory professor derya unutmaz leveraged gpt-5 pro to solve a three-year-old puzzle regarding how glucose metabolism affects t cell specialization, a mechanism central to cancer and autoimmune research. the model analyzed the data and correctly hypothesized that deoxyglucose inhibits il-2 protein construction, removing a barrier that usually prevents t cells from becoming inflammatory th17 cells. unlike previous versions, gpt-5 demonstrated the ability to synthesize complex biological context and identify non-obvious causal links, effectively functioning as an augmented scientific partner. this accelerates the shift of ai from a generative tool to a research engine capable of unstalling scientific progress in fields requiring deep domain expertise. success in highly specialized biology suggests future models will become standard infrastructure for interpreting complex experimental data.