Israeli researchers at Hebrew University announced that a machine‑learning framework sifted through more than 50 million chemical molecules to identify 92 promising candidates for future sugar substitutes. The AI‐powered method predicted sweet and bitter taste profiles and simulated how the candidates might bind to the human sweet‑taste receptor, offering a new way to discover low‑calorie sweeteners while avoiding unwanted aftertastes.
The team first compiled predictive models based on known sweet and bitter compounds. These models could distinguish between desirable sweetness and undesirable bitterness, which has long challenged food scientists. Using the models, the researchers screened three large chemical databases containing roughly 50.7 million molecules, filtering out those predicted to taste bitter.
After the initial filtering, the investigators further refined the list with machine‑learning algorithms and molecular docking techniques. Molecular docking evaluates how a small molecule, such as a potential sweetener, fits into a target protein—in this case, the single human sweet‑taste receptor. The detailed structure of the receptor’s component proteins was released the previous year, providing a template for the AI simulations.
This multistage approach reduced the candidate pool from the initial 50 million compounds to 556 molecules with favorable predicted properties. From those, the researchers selected 92 compounds for their predicted sweetness, lack of bitterness, and suitable binding to the sweet‑taste receptor. Many of the selected structures differ chemically from existing sweeteners, suggesting the possibility of entirely new classes of sugar alternatives.
Defining the sweet‑taste receptor as a target was a key step. The receptor, a protein on the surface of tongue cells, detects chemical signals and translates them into the perception of sweetness. By modeling how candidate molecules interact with this receptor, the AI team could anticipate which compounds might elicit a sweet taste in humans.
The study, published in the journal Food Chemistry, highlights how high‑throughput AI screening can accelerate the discovery of functional foods. The researchers noted that the method could identify potential sweeteners with properties not yet seen in commercial products, such as reduced calorie content or improved taste profiles.
Future work will involve laboratory synthesis and sensory testing of the 92 candidates to confirm the AI predictions. If successful, the approach could reduce the time and cost associated with bringing new sugar alternatives to market, addressing consumer demand for healthier sweetening options.
