Global Mid-Infrared Prediction Models Facilitate Simultaneous Analysis of Juice Composition from Berries of Actinidia, Ribes, Rubus and Vaccinium Species

Abstract

Introduction of Fourier-transform infrared (FTIR) spectroscopy would enable breeders to screen phenotypic variability in multiple fruit from large numbers of progeny. Thus far, however, there has been no comprehensive attempt to develop chemometric models for determination of quality attributes of small berry species by this approach. FTIR spectra (1800–900 cm−1) of juice from breeding populations of four genera (Actinidia, specifically red kiwifruit, Ribes, Rubus and Vaccinium) were analysed by partial least squares regression to determine the possibility of measuring soluble solids (SS), titratable acidity (TA) and total anthocyanin (ACY) concentration simultaneously using global prediction models. SS, TA and ACY concentrations across all berry juices ranged between 4.1 and 22.4 °Brix, 0.1–5.5% citric acid and 2–4697 ppm, respectively. R2 (coefficient of determination in cross-validation) and SECV (standard error of cross-validation) statistics for global models were 0.996 (0.22 °Brix), 0.996 (0.08% citric acid) and 0.893 (280 ppm). Analysis of data sets for individual berry types separately demonstrated that it was possible to develop models with superior prediction statistics for each attribute. However, these were not necessarily robust when validated against data from different seasons, locations or breeding selections. These global models represent an advance for researchers wishing to screen substantial fruit populations more rapidly.