External calibration models for the measurement of tomato carotenoids by infrared spectroscopy

A new method for the rapid construction of calibration models for the measurement of carotenoids in tomato by infrared spectroscopy was developed using a low-carotenoid tomato variety. Different amounts of lycopene and β-carotene were spiked in low-carotenoid tomato juice and homogeneously dispersed using a continuous stream of nitrogen and agitation. The protocol allowed for the production of homogeneous tomato juice standards with ∼90% carotenoid retention as determined by HPLC and UV–vis spectroscopy. Partial least-squares regression (PLSR) was used to create calibration models correlating infrared spectra from tomato lipid fractions and puree to their carotenoid content. The best calibration performance was obtained using lipid fraction models. PLSR calibration produced a high correlation coefficient (0.99) and a standard error of cross-validation of 0.81 mg/100 g for lycopene in the 0.0–14.0 mg/100 g range. Similarly, a high correlation coefficient (0.99) was obtained for β-carotene in the 0.0–9.0 mg/100 g range with a standard error of cross-validation of 0.36 mg/100 g. Calibration models were evaluated by predicting lycopene content in genetically diverse tomatoes from local markets. Results correlated well with concentrations determined by HPLC. This approach permitted the rapid preparation (∼3 min) of highly reproducible tomato juice standards that can be customized as needed to construct external calibration curves for the high-throughput measurement of carotenoids by FTIR, and potentially by other colorimetric and spectrometric techniques. The method eliminates the need for a large number of tomato samples for calibration and allows UV–vis spectroscopy to be used as an alternative reference method to HPLC.