TastePeptides-EEG: An Ensemble Model for Umami Taste Evaluation Based on Electroencephalogram and Machine Learning

Area of Science:

• Neuroscience
• Food Science
• Biotechnology

Background:

• Traditional food sensory evaluation relies on subjective human assessments, which can be inconsistent and prone to bias.
• Objective and reliable methods are needed to analyze taste perception, particularly for complex tastes like umami.

Purpose of the Study:

• To develop an electroencephalography (EEG)-based method for objective identification of umami substances.
• To establish a machine learning model for analyzing neural responses to umami taste.
• To investigate the neural correlates of umami taste perception.

Main Methods:

• Feature extraction from EEG data using percentage conversion, standardization, and significance screening.
• Selection of top-performing binary classification algorithms as submodels.
• Ensemble modeling using a Support Vector Machine (SVM) to integrate submodel outputs, creating the TastePeptides-EEG model.

Main Results:

• The TastePeptides-EEG model demonstrated high accuracy, with 90.2% on the validation set and 77.8% on the test set.
• Identified specific frequency changes in alpha (α) waves associated with umami taste perception.
• Discovered a novel frequency response delay phenomenon in the F/RT/C brain areas during umami taste stimulation.

Conclusions:

• EEG-based analysis provides an objective and effective method for identifying umami substances, overcoming limitations of manual sensory evaluation.
• The study reveals new insights into the neural mechanisms underlying umami taste perception.
• The publicly available TastePeptides-EEG model facilitates research in flavor perception and brain-computer interfaces for taste analysis.