In brief
A team at Hebrew University used AI to comb through 50 million compounds, filtering for sweet taste and avoiding bitterness. They simulated receptor binding and narrowed the list to 92 candidates, many with novel chemical structures that could launch new classes of sugar alternatives. The results, published in *Food Chemistry*, highlight AI’s role in food innovation.
According to Xinhua, researchers at Hebrew University in Jerusalem announced that their new machine‑learning framework has identified 92 promising sugar‑substitute candidates after scanning more than 50 million chemical molecules.
The AI model was built from known sweet‑ and bitter‑tasting compounds, allowing it to predict desirable taste profiles and eliminate molecules likely to cause bitterness.
It was then applied to three large databases that together contain about 50.7 million molecules; the preliminary screening yielded 556 potential sweeteners, which were narrowed to 92 after additional machine‑learning refinement and molecular docking simulations that examined how each candidate might bind to the single human sweet‑taste receptor.
Many of the selected compounds have chemical structures that differ from those of existing sweeteners, potentially opening the door to entirely new classes of low‑calorie sugar alternatives.
The findings were published in the journal Food Chemistry, where the team highlighted AI and molecular modeling as tools to accelerate the discovery of sweeteners while minimizing unwanted bitter or metallic aftertastes.
Scientists plan to advance the most promising candidates through laboratory testing to confirm their safety, efficacy, and sensory qualities.
