Modelling Through Artificial Neural Networks of the Phenolic Compounds and Antioxidant Activity of Blueberries

Abstract

The present study aimed at investigating the influence of several production factors, conservation conditions, and extraction procedures on the phenolic compounds and antioxidant activity of blueberries from different cultivars. The experimental data was used to train artificial neural networks, using a feed-forward model, which gave information about the variables affecting the antioxidant activity and the concentration of phenolic compounds in blueberries. The ANN input variables were location, cultivar, the age of the bushes, the altitude of the farm, production mode, state, storage time, type of extract and order of extract, while the output variables were total phenolic compounds, tannins as well as ABTS and DPPH antioxidant activity. The ANN model was fairly good, with values of the correlation factor for the whole dataset varying from 0.948 to 0.979, while the values of mean squared error were ranging from 0.846 to 0.018, for DPPH antioxidant acidity and anthocyanins, respectively. The results obtained showed that the methanol extracts contained higher amounts of total phenols and anthocyanins as compared to acetone: water extracts, while presenting similar quantities of tannins in both extracts. The blueberries from organic farming were richer in phenolic compounds and possessed higher antioxidant activity than those from conventional agriculture. Even though the effect of storage was not established with high certainty, a trend was observed for an increase in the phenolic compounds and antioxidant activity along storage, either when under refrigeration or under freezing, for the storage periods evaluated.