Beyond single-marker analysis: integrating FTIR fingerprinting for the pre-bottling assessment of light-strike susceptibility

Abstract

Light-strike susceptibility varies widely among rosé wines, yet predictive tools for pre-bottling risk assessment remain inadequate. While the photo-sensitizing role of riboflavin and the precursor function of methionine are well-established, quantification of individual markers often fails to capture the complexity of wine matrix interactions. This study aimed to develop robust models to estimate methanethiol (MetSH) formation by integrating routine enological parameters with FTIR spectroscopy. Thirty-six commercial Chiaretto di Bardolino wines (2023 vintage) were collected at the bottling line. Initial composition was characterized by routine assays, HPLC (riboflavin), LC-MS (methionine), and SPME-GC-MS (VSCs). FTIR spectra (900–3000 cm^–1) were acquired to provide a holistic fingerprint of the wine matrix. After anoxic stabilization, aliquots were exposed to LED light (400–500 nm) for 30 days at 20 °C. To mitigate the risk of overfitting inherent to high-dimensional spectral data, predictive modeling was performed using Ridge regression coupled with bootstrap selection (2000 iterations), ensuring the identification of stable and generalizable predictors. All wines accumulated MetSH upon exposure, but with a striking 5-fold range in net production (1.9-10.6 µg/L). A chemistry-only model R2 = 0.65 (LOOCV) identified key drivers including riboflavin, methionine, and total acidity (positive coefficients), and free/total SO₂ (negative). Integrating FTIR second-derivative features significantly improved the prediction R2 = 0.72 (LOOCV), with selected regions (O-H/N-H stretching, C-O vibrations) likely capturing the “compositional buffering capacity” and antioxidant network of the wine. Notably, a simplified model, independent of HPLC/LC-MS data, maintained similar performances R2 = 0.67 (LOOCV) using only routine winery parameters and FTIR spectra. These results demonstrate that light-strike risk is a matrix-dependent property that can be effectively modeled. This approach provides a practical, non-destructive screening tool for immediate winery implementation, enabling producers to make evidence-based decisions on glass color, SO₂ management, and shelf-life strategies before the wine leaves the cellar.

References

Fracassetti D., Di Canito A., Bodon R., Messina N., Vigentini I., Foschino R., Tirelli A., Trends Food Sci. Technol., 2021, 112, 547–558.