Predicting and Co-Optimizing the Taste and Aroma of Green Tea During Spreading Using the TabPFN Model

Area of Science:

• Food Science
• Agricultural Engineering
• Analytical Chemistry

Background:

• Spreading conditions during green tea processing critically influence sensory attributes.
• Understanding the impact of dehydration-induced mechanical stress on cellular changes is essential for quality control.

Purpose of the Study:

• To investigate the effects of spreading conditions on green tea taste and aroma.
• To develop a generalizable prediction model using small data for process optimization.
• To quantify the relationship between spreading parameters and key taste/aroma metabolism.

Main Methods:

• Integration of Scanning Electron Microscopy (SEM), non-targeted dual-omics, and TabPFN ( a machine learning model).
• Central composite design for experimental analysis.
• Establishment of chemical-sensory proxies: relative polyphenol-to-amino acid ratio (R-PAR) and floral intensity index (FII) using UHPLC-HRMS and HS-SPME-GC-MS.

Main Results:

• Dehydration stress alters cell membrane permeability, affecting taste and aroma compound release.
• Optimized conditions (76.8% initial moisture, 26.2°C, 61.5% RH, 0.85 m/s air speed) yielded R-PAR of 0.465 and FII of 125.70.
• TabPFN demonstrated superior prediction accuracy (R² of 0.81 and 0.77) compared to RSM, PLSR, and SVR on small-sample datasets.

Conclusions:

• TabPFN is suitable for small-sample tea processing modeling and intelligent optimization.
• The study quantifies the link between spreading conditions and green tea's sensory profile.
• Provides a methodological basis for digital precision in green tea production.