A-TEEM and machine learning for rapid assessment of grape anthocyanins and wine smoke taint

Abstract

Anthocyanin profiling remains essential for assessing red grape quality and predicting wine color stability. However, traditional HPLC-based methods are time-consuming and impractical for routine vineyard maturity monitoring. Hybrid grapes present additional challenges due to their distinct anthocyanin composition, including diglucoside forms for which reference standards are difficult to obtain and costly. A-TEEM spectroscopy, which simultaneously captures absorbance and three-dimensional fluorescence excitation-emission matrices, presents a promising alternative for rapid phenolic characterization without requiring expensive standards. Wildfire smoke exposure poses an escalating threat to wine quality globally, yet rapid analytical methods for identifying affected fruit and wine remain unavailable. Current smoke taint detection relies on GC-MS and LC-MS/MS approaches requiring 24–48 hours and extensive sample preparation, making real-time assessment impractical. This work develops A-TEEM-based methods coupled with machine learning for rapid assessment of both hybrid grape anthocyanins and smoke taint in wine. For anthocyanin profiling, grape samples were analyzed by HPLC-DAD to establish reference anthocyanin compositions. A-TEEM spectral data combined with HPLC reference measurements were used to build machine learning models for quantitative prediction of anthocyanins. For smoke taint assessment, wine samples from smoke-impacted 2025 vintage fruit were analyzed by LC-MS/MS to determine volatile phenol glucoconjugates reference chemistry. A-TEEM spectral data combined with LC-MS/MS reference measurements were used to build machine learning models for quantitative prediction of smoke taint markers. Results demonstrate that A-TEEM combined with machine learning provides rapid quantification of both hybrid grape anthocyanins and smoke taint compounds, reducing analysis time from conventional analytical methods to minutes. This approach enables timely vineyard and winery decisions and addresses analytical challenges in hybrid viticulture and wildfire impact assessment.

References

Szeto, C.; Feng, H.; Sui, Q.; Blair, B.; Mayfield, S.; Pan, B.; Wilkinson, K. American Journal of Enology and Viticulture. 2024 75:0750013; DOI: 10.5344/ajev.2024.23060