Among the multiple perceptions triggered by the gustatory system, bitterness is usually associated with the avoidance of food and is believed to have evolved to alert us against the consumption of toxic plants. The human genome possesses 25 functional TAS2R genes encoding bitter taste receptors that are differentially activated by a broad range of chemically and structurally diverse bitter compounds. TAS2Rs belong to the G protein-coupled receptor (GPCR) family, and while several GPCR structures have been experimentally solved, the exact tridimensional structure of TAS2Rs has yet to be determined. Without such key structural information, predicting the activity and mechanism of action of bitter molecules on TAS2Rs mostly relies on molecular modeling. Here, we present an integrative computational protocol that combines sequence alignment, homology modeling and constraints derived from site-directed mutagenesis data to build relevant 3D models of all human TAS2Rs. Using TAS2R16 as a test case, we challenged the accuracy of our model by mutating the positions we identified as functional molecular switches involved in ligand sensing and downstream signaling. These results provide molecular insights on structure-function relationships of bitter taste receptors that could be extended to the mammalian repertoire, and layout the groundwork for the development of bitter taste modulators. Funding: French Ministry of Higher Education and Research, UCAJEDI “Investments in the Future” grant number ANR-15-IDEX-01, National Research Foundation of Korea grant number NRF2020R1A2C2004661, GIRACT (Geneva, Switzerland), the Gen Foundation (Registered UK Charity No. 1071026), and support from the Université Côte d’Azur’s Center for High-Performance Computing.