In brief
Hebrew University researchers developed an AI system that screened more than 50 million molecules, predicting sweet and bitter taste profiles and simulated binding to the human sweet‑taste receptor. The study, published in *Food Chemistry*, narrowed the list to 92 unique candidates with structures unlike existing sweeteners, paving the way for next‑generation low‑calorie sweeteners.
Hebrew University researchers have announced that a machine‑learning framework identified 92 promising sugar‑alternative candidates from a pool of more than 50 million chemical compounds.
The study, published in the journal Food Chemistry and reported by Xinhua, used predictive models built on known sweet and bitter molecules to sift through three chemical databases with approximately 50.7 million entries, flagging compounds predicted to be sweet while discarding those likely to taste bitter.
Following the initial screen, the team employed machine‑learning combined with molecular docking to simulate how the shortlisted molecules might bind to the single human sweet‑taste receptor, whose component protein structure was released last year. The refinement process narrowed the pool to 556 potential sweeteners, from which 92 candidates were chosen for their favorable predicted properties and suitability for future testing.
According to Xinhua, many of the selected compounds feature chemical frameworks that differ from those of existing sweeteners, potentially opening new classes of sugar alternatives.
