Texas A&M Health Institute of Biosciences and Technology researchers announced on June 8 the development of an artificial intelligence-designed molecular switch that uses caffeine to control engineered proteins inside living cells. The platform, called CODS (caffeine-operated dissociation system), aims to help scientists build safer and more controllable gene and cell therapies.
The study, published in the Journal of the American Chemical Society, was led by Yubin Zhou, MD, PhD, FAAAS, FAIMBE, FRSC. Graduate students Brendan McKee and Tatsuki Nonomura played central roles in driving the AI-guided protein design and conducting live-cell validation studies. The new technology builds on previous caffeine-responsive systems but differs by using caffeine to separate rather than join engineered proteins—a distinction important for therapies needing not only activation but also ways to pause or reset cellular activity.
To create CODS, the team used AI-guided protein design to engineer a synthetic binder that recognizes a caffeine-responsive protein module. Without caffeine present, the binder holds proteins together; when caffeine is added, it causes them to separate. "Many genetically-encoded molecular tools act like accelerators," Tianlu Wang said. "CODS gives us something closer to a brake or pause button." The project relied heavily on Texas A&M High Performance Research Computing services for computational modeling.
The researchers demonstrated CODS in three applications: controlling gene activity by reducing it upon adding caffeine; regulating programmed cell death through a system where caffeine triggers inflammatory cell death; and modulating CAR T-cell therapies so that CAR T-cells can be temporarily deactivated with caffeine—potentially serving as a safety switch during cancer immunotherapy treatments.
Zhou clarified that while coffee itself is not medicine, “caffeine can help us imagine medicines that are more controllable, more responsive and safer for patients.” He added that similar strategies could eventually use other familiar molecules as switches for therapeutic purposes.
Further testing will be required before clinical application of CODS in therapeutic cells or disease models. “Powerful therapies need powerful control,” Zhou said. “By combining AI-designed proteins, high-performance computing and familiar small molecules, we are building a new language for communicating with engineered cells.”
