AI-designed RNA switch enables logical NAND function in living cells

Researchers at the Technical University of Darmstadt have developed an RNA-based genetic switch that behaves like a digital NAND gate. Designed using artificial intelligence and laboratory screening, the system enables cells to make logical decisions about gene expression.

The findings were published in the journal Nucleic Acids Research. According to the university, the approach combines synthetic biology with artificial intelligence to design new RNA elements capable of controlling gene expression in living cells.

The RNA switches are based on riboswitches – RNA sequences that respond to small molecules known as ligands. When these ligands bind to the RNA structure, the molecule changes its shape. This structural change blocks the ribosome, the cellular machinery responsible for producing proteins, thereby regulating translation.

Riboswitches offer several advantages for synthetic biology:

• They function without additional proteins
• They are extremely small, typically fewer than 100 nucleotides
• They require little cellular energy to produce
• They place minimal metabolic burden on the cell

Dr. Daniel Kelvin from the Centre for Synthetic Biology at TU Darmstadt demonstrated that connecting two riboswitches in a continuous RNA sequence can create dual-input genetic switches. "We use these RNA-based dual-input switches to implement logical functions in living cells, similar to those in computers. To achieve this, we have constructed a combination of two riboswitches that functions like a Boolean NAND gate."

In digital electronics, a NAND gate outputs an "off" signal only when both inputs are "on" In all other cases, the signal remains "on". Translated into a biological system, this means that gene expression is turned off only when two specific ligands bind simultaneously to the riboswitch. If one or both ligands are absent, the gene remains active.

Such complex behavior has not been observed naturally. Designing the hybrid riboswitch therefore required extensive experimentation and computational modeling because the number of possible RNA sequence combinations increases rapidly with sequence length.

The researchers combined laboratory screening with Bayesian optimization, a machine-learning method, to design improved RNA variants. First, they constructed a hybrid riboswitch that showed partial NAND-like behavior. This served as the basis for creating a large library of RNA variants.

Thousands of variants were generated, particularly in the central communication module connecting the two ligand-binding regions. The variants were tested using flow cytometry, which allowed precise measurement of their behavior under different ligand combinations.

To improve efficiency, the researchers used the Kriging Believer method, which allows several candidate RNA sequences to be proposed simultaneously rather than sequentially. This enables multiple experiments to run in parallel and prevents the algorithm from selecting overly similar sequence variants.

After evaluating only 82 variants, the system identified several optimized riboswitches. The best candidate displayed a nearly digital NAND behavior, with a clear separation between "on" and "off" states.

The development of a functional NAND riboswitch is significant because NAND gates can be used to build all other logical functions, including AND, OR, and XOR. This capability opens new possibilities for programming biological systems.

Cells could potentially be engineered to make logical decisions – for example, producing a specific compound only when certain combinations of nutrients or signaling molecules are present.

Possible applications include:

• biosensors for detecting metabolic states
• diagnostics that recognize tumor-specific molecular patterns
• environmental monitoring systems that respond to combinations of toxins

MEDICA-tradefair.com; Source: Technische Universität Darmstadt