Flavonoids are a class of bioactive compounds extremely important in food and wine industry. The development of rapid methods for their quantification in grape berries is one of the modern challenges in viticulture and enology research. Total flavonoid (TF) amount changes during grape ripening and also berry physicalmechanical properties, as evaluated by instrumental texture analysis, change in the same period. In this work, TF and berry physical-mechanical parameters were linked together through predictive models. Models were developed for each of four red wine grape cultivars: Brancellao, Cabernet Franc, Mencía and Merenzao, and another one considered all cultivars together. These models reached high accuracy and allowed to predict TF in grape berries with a low error (RMSE from 0.15±0.07mgg-1 to 0.35±0.10mgg-1 in prediction, as evaluated by cross-validation).Berry weight (BW) was the parameter having the largest influence on TF predictions, and also was the only variable having part in all models. BW and chewiness had a similar behavior and when berry weight was excluded, chewiness was able to substitute its role in all models. The other physical-mechanical characteristics displayed a different behavior across cultivars.In conclusion, this work shows that it is possible to predict TF from physical-mechanical predictors in grape berries and that cultivar specific models reach higher accuracy for this purpose than the multi-cultivar model.
- Berry weight
- Mechanical properties
- Multivariate adaptive regression splines (MARS)
- Red wine grape cultivars
- total flavonoids