Tidsskrift: Molecular Nutrition & Food Research, vol. 50, p. 811–817, 2006
Open Access: none
In order to study the formation of acrylamide in potato crisps during processing, an experimental design was set up. The design variables were drying time (6 levels), frying temperature (2 levels) and frying time (8 levels). The design contained 36 samples, which were analysed for acrylamide contents using LC high-resolution mass spectroscopy (LC-HRMS), and fat contents using the Soxhlet apparatus. Prior to analysis, all potato crisp samples were ground and analysed on an NIRSystems 6500 near-infrared (NIR) spectrometer. The acrylamide contents were modelled by: (i) design variables using multiple linear regression, (ii) NIR spectra using partial least squares regression (PLSR) and (iii) design variables and NIR spectra in combination using a novel technique combining least squares regression on the former, and PLSR on the latter. The results showed that the NIR spectra alone or in combination with the design variables gave better prediction models for acrylamide than the design variables alone. This implies that the spectra contain chemical information that is not purely a result of the processing variables that were investigated in this experiment. NIR spectroscopy is proposed as a possible tool for screening and identification of potato crisps with a high acrylamide content.