Machine learning approaches to predict TAS2R receptors for bitterants

Area of Science:

• * Sensory science and computational biology.
• * Explores the intersection of taste receptor research and artificial intelligence.

Background:

• * Bitter taste perception relies on taste receptor type 2 (TAS2R) G protein-coupled receptors, implicated in toxin detection, glucose homeostasis, and immune responses.
• * Human TAS2Rs exhibit polymorphism, varying in localization and function, leading to diverse signaling pathway activation based on tissue and ligand.
• * In vitro screening for TAS2R ligands is resource-intensive, highlighting the need for efficient in silico prediction methods.

Purpose of the Study:

• * To provide an overview of machine learning (ML) based in silico methods for predicting bitterant-TAS2R interactions.
• * To critically evaluate existing ML models (BitterX, BitterSweet, BitterMatch) for their ability to predict specific receptor-ligand associations.
• * To identify limitations and establish a foundation for future research in this domain.

Main Methods:

• * Literature review and analysis of state-of-the-art ML-based models for bitterant-TAS2R interaction prediction.
• * Focus on three specific ML models: BitterX (2016), BitterSweet (2019), and BitterMatch (2022).
• * Assessment of data availability and challenges in training receptor-ligand association models.

Main Results:

• * Numerous ML-based taste classifiers exist, primarily for bitter/non-bitter or bitter/sweet classification.
• * Few ML models specifically predict which TAS2R receptors are targeted by bitter molecules due to data scarcity and incompleteness.
• * The reviewed models (BitterX, BitterSweet, BitterMatch) represent key advancements but have limitations.

Conclusions:

• * In silico prediction of bitterant-TAS2R interactions is crucial for efficient ligand and target selection in experimental studies.
• * Existing ML models offer valuable tools but require further development to overcome data limitations.
• * This review sets the stage for future research aimed at improving the accuracy and scope of bitterant-TAS2R prediction models.