Comprehensive flavor profiling and neural perception of yeast extracts from five yeast sources via sensomics, machine learning, and EEG.

Yeast extract is widely used as a natural flavor enhancer, yet how yeast source shapes its taste, aroma, and neural perception remains insufficiently understood. Here, we compared yeast extracts from five industrially relevant sources (beer yeast, BY; bread yeast, BRY; Candida utilis, CUY; alcohol yeast, AY; wine yeast, WY) using an integrated sensory-chemical-neural framework. Sensory evaluation revealed clear source-dependent flavor signatures (p < 0.05), with CUY showing the strongest umami intensity. Solvent-assisted flavor evaporation (SAFE) coupled with GC-O-MS identified 120 volatile compounds, of which 52 key odorants (OAV ≥ 1, FD ≥ 32) were validated by aroma recombination (similarity 8.3/10) and omission tests. Pearson correlation linked precursor amino acids with key aroma compounds through Strecker degradation and Maillard reaction pathways. Exploratory machine learning classification achieved cross-validation accuracies of 96.8% (e-nose) and 90.8% (key aroma), with independent validation accuracies of 86.7% and 80.0%, respectively. EEG analysis revealed distinct neural perception patterns consistent with sensory evaluation. These findings establish a sensory-chemical-neural framework for evaluating yeast extract flavor quality and guiding source-specific food flavor design.