The analysis of carotenoids is complicated by the tendency of these compounds to react with radical species, leading to oxidative breakdown and isomerization during extraction. Therefore, protocols should be rapid and avoid unnecessary exposure to heat, acids, and so forth. Here, we evaluate the use of visible and near infrared reflectance spectroscopy (Vis/NIRS) to measure carotenoid contents in fruit from 28 Musa (banana and plantain) varieties. Carotenoid contents were first quantified using standardized RP-HPLC protocols, and these results were then used to develop algorithms to predict carotenoid contents from Vis/NIR spectra of the same samples. Cross-validation of the predictive algorithms across a genetically diverse group of varieties demonstrated that correlation coefficients between the HPLC measurements and the Vis/NIRS predictions varied from good for the total carotenoids and β-carotene fractions (r2cv, 0.84, 0.89) to reasonable for α-carotene and cis-carotenes (r2cv, 0.61, 0.66), but there was only a poor correlation (r2cv, 0.30) for the minor lutein component. Nonetheless, since 90% of the Musa carotenoids consist of only α- and β-carotene, results indicate that Vis/NIRS can be used for the high-throughput screening of fruit pulp samples for vitamin A nutritional content on the basis of their total carotenoids content.
Journal of Agricultural and Food Chemistry (2009) 57 (5) 1742-1751 [doi: 10.1021/jf803137d]