How can cells be programmed to respond to complex signals and make targeted decisions, similar to a logical circuit in a computer? An interdisciplinary team from two research groups at the Centre for Synthetic Biology at TU Darmstadt has developed a new approach: an RNA-based genetic switch. The results were published in the journal Nucleic Acids Research.

The switch is based on so-called riboswitches: short sections of messenger RNA (mRNA) that can respond to specific small molecules (“ligands”). When the ligand binds to the “switch,” the shape of the RNA changes, and the ribosome – which would normally produce a protein following the instructions encoded in the mRNA – is blocked.

Riboswitches are particularly attractive for synthetic biology because they function without additional proteins, are very small, and require very little energy for their production in the cell. This makes them ideal tools for synthetic gene regulation.

The TU Darmstadt research team has now combined two such riboswitches. The result is a switch capable of evaluating two different molecular signals simultaneously. First author Dr. Daniel Kelvin, a researcher at the Centre for Synthetic Biology at TU Darmstadt, demonstrated that seamlessly linking two riboswitches enables the creation of genetic switching elements with two different inputs.

Computer ‘functions’ in living cells

“We use these RNA-based dual-input switches to implement logical functions similar to those in computers in living cells,” says Kelvin. “To do this, we constructed a combination of two riboswitches that functions like a Boolean NAND gate.”

A NAND gate is a fundamental component of digital electronics. It produces an “off” signal only when both inputs are active at the same time. In all other cases, the signal remains “on.”

Transferred to biology, this means that gene expression – the production of proteins encoded in genes – is switched off only when two different ligands bind to the riboswitch simultaneously. If even one of the two ligands is missing, the gene remains active. Such behavior is complex and has not previously been observed in nature. In addition, the number of possible sequence variants grows exponentially with sequence length.

Lab experiments combined with AI

This made the construction of the hybrid NAND riboswitch a major challenge. To identify suitable variants, the team combined laboratory experiments with methods from artificial intelligence. The researchers first generated thousands of variants of the RNA switch. They then tested in the laboratory how these variants responded to different combinations of ligands. The results served as training data for a computer program.

Erik Kubaczka, also a researcher at the Centre for Synthetic Biology and co-author of the publication, explains: “A deep learning model then predicts which RNA variants best perform the NAND function. Our optimization algorithm, based on Bayesian optimization, then selectively chooses new candidates – and learns from each experiment.”

Using this approach, the team was able to identify several strongly improved RNA switches after testing only 82 variants. The best candidate showed a very clear separation between the “on” and “off” states.

Biosensors for medicine and environmental monitoring

With the new hybrid riboswitch and the AI-based design approach, the team led by TU Professor Beatrix Süß (Centre for Synthetic Biology, Synthetic RNA Biology group) and Professor Heinz Koeppl (Centre for Synthetic Biology, Self-Organizing Systems group) provides a way to design biological circuits in a more targeted manner. Since many other logical functions can be constructed from NAND gates, living cells could in the future learn to make more complex decisions – for example, producing a substance only when specific combinations of nutrients or signaling molecules are present.

This could also enable the development of biosensors for medicine and environmental monitoring – for instance, sensors that detect certain metabolic states, identify tumor signatures, or report environmental toxins in specific combinations.

The project illustrates how biology and artificial intelligence are increasingly converging – and how machine learning helps discover new functional RNA elements that nature itself has never produced.

Reference: Kelvin D, Kubaczka E, Karava M, Koeppl H, Suess B. Iterative design of a NAND hybrid riboswitch by deep batch Bayesian optimization. Nucleic Acids Research. 2026;54(5):gkag145. doi: 10.1093/nar/gkag145

This article has been republished from the following materials. Note: material may have been edited for length and content. For further information, please contact the cited source. Our press release publishing policy can be accessed here.