Designing Novel Sweetener Candidates Targeting the Human Sweet Taste Receptor via a Target-Aware Chemical Language Model

Excessive intake of dietary sugar is a leading, modifiable risk factor for non-communicable diseases. This drives a critical need for safe sugar alternatives. Widespread artificial sweeteners pose health risks and limitations. This study aimed to computationally design novel small-molecule sweetener candidates targeting the human Venus flytrap domain (VFTD) of the sweet taste receptor (TAS1R2-TAS1R3) and evaluate their binding affinity, ADMET properties, sweetness intensity, and thermal stability. An AI-driven pipeline with a target-aware chemical language model (TamGen) generated two molecular libraries. The binding affinity of each generated compound to the VFTD of the TAS1R2 subunit was evaluated using molecular docking simulations with AutoDock Vina. Sucralose was used as a benchmark cutoff, and compounds with higher binding affinity than sucralose (< -5.5 kcal/mol) were selected. The selected compounds were subjected to in silico ADMET profiling using ADMETlab 3.0 to assess toxicity, pharmacokinetics, and synthetic accessibility scores. Finally, sweetness intensity (logSw) and thermal stability were predicted using Chemsweet software. TamGen framework generated 19,658 unique molecules. Initial docking-based screening selected 18,454 compounds having higher binding affinity than sucralose. Subsequent ADMET profiling further reduced this library to 17 compounds. These compounds showed a very favorable safety profile, including low toxicity and poor intestinal absorption. In addition, these selected compounds show logSw values between 4.08 and 5.63, indicating higher sweetness intensity than sucrose. Thermal stability analysis has confirmed that these compounds are highly stable and suitable for food processing applications. The integration of generative AI models and rigorous computational evaluation provides an efficient way to discover next-generation sweeteners. The identified candidates have a high sweetness potential, excellent thermal stability, and favorable safety profiles. To validate these computational results, future research needs to integrate in vitro receptor-binding assays and sensory evaluations, alongside in vivo toxicological studies, to confirm overall physiological safety and metabolic effects.