Summary: Researchers applied AlphaFold3, the latest AI-driven protein structure prediction system, to model the three-dimensional structures of all 25 known human bitter taste receptors (T2Rs). When benchmarked against experimental data and previous AlphaFold2 predictions, AlphaFold3 produced consistently more accurate structural models.
The analysis revealed strong conservation of intracellular regions across T2Rs and notable variation in their extracellular domains, which helps explain how different receptors recognize a wide range of bitter compounds. These structural insights emphasize T2Rs’ roles both in taste perception and gut–brain signaling, suggesting new directions for research into appetite control, glucose regulation, and lifestyle-related conditions such as diabetes.
Key Facts
- AI Upgrade: AlphaFold3 outperforms AlphaFold2 in predicting T2R structures.
- Structural Insights: Intracellular regions are highly conserved; extracellular regions display extensive variation.
- Health Potential: Results link bitter receptors to gut–brain signaling and suggest potential relevance for metabolic disease research.
Source: Shibaura Institute of Technology
Receptor proteins—whether displayed on the cell surface or located within the cell—bind signaling molecules called ligands and trigger downstream cellular responses. Taste receptors, in particular, are specialized to detect tastants, the chemical compounds that produce taste sensations.
Bitter taste receptors (T2Rs) mediate the sensation of bitterness in the mouth, but they are also expressed outside the oral cavity. Notably, T2Rs appear in neuropod cells of the gastrointestinal tract, which transmit sensory information from the gut to the brain. This extraoral expression links T2Rs to functions of the gut–brain axis and suggests roles in physiological processes such as glucose homeostasis and appetite regulation.
To date, 25 human T2R subtypes have been identified, yet experimentally resolved structures exist for only two: T2R14 and T2R46. Advances in AI-based modeling have enabled increasingly accurate prediction of protein structures, prompting researchers to re-evaluate T2R architecture with the latest tools.
Earlier work used AlphaFold2 (AF2), a high-profile AI model, to predict T2R structures. With the release of AlphaFold3 (AF3), the research team led by Professor Naomi Osakabe at Shibaura Institute of Technology reassessed all 25 human T2Rs to compare AF3 predictions with AF2 outputs and with the available experimental structures.
“Because bitter taste receptors are present in the gastrointestinal tract, they likely influence the gut–brain axis, glucose tolerance, and appetite,” said Prof. Osakabe. “Accurate structural models are essential to clarify how these receptors function in both taste and metabolic signaling.”
The study, published online July 14, 2025 and appearing in Volume 11 of Current Research in Food Science on July 22, 2025, was authored by Takafumi Shimizu and Rio Ohno (Shibaura Institute of Technology) together with Prof. Vittorio Calabrese (University of Catania).
Researchers retrieved amino acid sequences for all human T2Rs from UniProt and used AlphaFold3 to generate three-dimensional structure predictions. For comparison, they accessed previously generated AlphaFold2 models from the AlphaFold database and obtained experimentally solved structures for T2R14 and T2R46 from the Protein Data Bank. A suite of structural visualization and analysis tools was applied to align models and quantify accuracy.
Benchmarking against 115 cryo-EM structures for T2R14 and three experimental structures for T2R46 demonstrated that AlphaFold3 models matched experimental data more closely than AlphaFold2 in all tested cases. Although local evaluation metrics such as predicted local distance difference test (pLDDT) scores were lower for AF3 across some T2R subtypes, global agreement with experimental structures and overall model quality favored AF3.
Structural comparisons showed a clear pattern: intracellular regions across T2Rs were structurally conserved, whereas extracellular regions—where ligand binding occurs—exhibited substantial diversity. Based on structural similarity and sequence identity, the team classified the T2Rs into three clusters, a grouping that may clarify why specific receptors respond to particular bitter molecules.
These structural features likely underlie the receptors’ ability to recognize thousands of chemically diverse bitter substances, often via interaction with the taste receptor–specific G protein α-gustducin. Given T2Rs’ presence in both oral and gastrointestinal tissues, the authors highlight potential implications for drug discovery and nutritional research, particularly strategies aimed at metabolic disease prevention and management.
The paper recommends further study of the relationship between T2R sequence variation, three-dimensional structure, and individual differences in taste perception. Such research will deepen understanding of receptor–ligand interactions and inform therapeutic and dietary approaches that target T2R-mediated pathways.
Funding Information: This work was supported by JSPS KAKENHI (Grant Number: 23H02166).
About this neuroscience research news
Author: Kohei Tsuchiya
Source: Shibaura Institute of Technology
Contact: Kohei Tsuchiya – Shibaura Institute of Technology
Image: The image is credited to Neuroscience News
Original Research: Open access.
“The three-dimensional structure prediction of human bitter taste receptor using the method of AlphaFold3” by Naomi Osakabe et al. Current Research in Food Science
Abstract
The three-dimensional structure prediction of human bitter taste receptor using the method of AlphaFold3
Bitter taste receptors (T2Rs), members of the G protein–coupled receptor family, are expressed in oral tissues and at extraoral sites where they participate in physiological pathways such as gut–brain communication. Despite their importance, experimentally resolved structural data are limited to two human T2Rs (T2R14 and T2R46).
This study evaluates AlphaFold3 (AF3) as a tool for predicting structures of all 25 human T2Rs and contrasts its performance with AlphaFold2 (AF2). AF3 models were validated against existing experimental structures and showed improved agreement with empirical data. While AF3 produced lower predicted local confidence scores across some receptor subtypes, its overall structural fidelity surpassed AF2.
Analysis revealed conserved intracellular architecture alongside diverse extracellular domains, consistent with the need to recognize many distinct bitter ligands. Clustering by sequence identity and root mean square deviation identified discrete receptor groups, suggesting structural determinants for ligand specificity. These findings support the use of AF3 to accelerate research on T2R biology and its implications for health, including metabolic conditions such as obesity and diabetes.