Yubin Zhou, MD, Ph.D., and his team at the Texas A&M Health Institute of Biosciences and Technology have engineered a caffeine-activated molecular switch that lets scientists pause, resume, and precisely control living cells on demand—a breakthrough that could transform how gene and cell therapies work inside the human body.
The platform, called CODS (caffeine-operated dissociation system), works like a molecular clasp. Without caffeine, synthetic proteins stay locked together, keeping cellular functions running. When caffeine enters the cell, the proteins separate, triggering a pause or reset. This matters because existing gene therapies mostly act like accelerators—they turn things on. Medicine often needs the opposite: a brake button. CODS provides exactly that.
The team used artificial intelligence to design the system from scratch. Rather than cobbling together proteins found in nature, they used AI-guided protein design algorithms and molecular simulations to create a small synthetic binder that recognizes caffeine-responsive protein modules. The Texas A&M High Performance Research Computing service provided the computational power to evaluate thousands of candidate designs before testing the most promising ones in living cells. The result: a system that responds to very low caffeine concentrations, works within minutes, and can be reversed repeatedly.
The researchers demonstrated CODS working three ways. First, they controlled gene activity—flipping a genetic switch on and off by adding or removing caffeine. Second, they rewired a cell-death protein with the caffeine-responsive switch, creating a system that could trigger programmed cell death when needed. This opens possibilities for studying inflammation and for designing therapeutic cells that can be eliminated if something goes wrong. Third, and most translationally significant, they tested CODS on CAR T-cells, the engineered immune cells that recognize and attack cancer.
Tianlu Wang, Ph.D., and colleagues, including graduate students Brendan McKee and Tatsuki Nonomura, drove the computational modeling and molecular engineering. McKee led the AI-guided protein design; Nonomura performed the live-cell validation studies. "Many genetically-encoded molecular tools act like accelerators," Wang said. "CODS gives us something closer to a brake or pause button." Zhou added, "AI is changing how we design biology. Instead of relying only on protein parts that already exist in nature, we can now design new mini proteins with specific behaviors. Here, we used AI to help turn caffeine into a precise trigger for controlling engineered cells."
The work, published in the Journal of the American Chemical Society, builds on Zhou's earlier caffeine-responsive technologies but represents a fundamental shift. Previous systems showed caffeine could bring proteins together; CODS does the opposite. That reversal is what makes it powerful for safety and control.
The practical implications are profound. Future cell therapies—especially CAR T-cell treatments—could be made far safer if physicians could pause or reset them at any moment simply by managing caffeine exposure. For patients who develop dangerous side effects, this offers a kill-switch. For researchers, it opens new windows into how engineered cells behave. Zhou's directorship of the Center for Translational Cancer Research positions this work squarely in the pipeline toward human trials. High-performance computing made the speed possible; AI made the design possible. Caffeine, that ancient stimulant, may soon help save lives.