Chemometric classification and quantification of cold pressed grape seed oil in blends with refined soybean oils using attenuated total reflectance-mid infrared (ATR-MIR) spectroscopy

WOS:000452946300018Of the cold pressed edible oils, grape seed oil (GSO) is one of the highest quality and, consequently the most expensive oil. Because of that, adulteration by mixing GSOs up with lower price edible oils is frequently seen. The classification and determination of adulterant concentration for GSO is hence a focus of great interest. The aim of this study was to investigate the feasibility of ATR MIR spectroscopy coupled with the multivariate methodology for qualitative and quantitative analyses of cold pressed GSO adulteration with refined soybean oil (SBO), by using characteristic wavenumber regions. Thirty three pure oils and ninety six blends were analyzed using ATR MIR spectroscopy in combination with PCA, WA, SIMCA and PLSR analysis. SIMCA models provided an excellent classification for pure GSO and other samples. The classification limits by ATR MIR spectroscopy was also below 5%. Quantitative analyses were performed by minimization of RMSE of cross-validation values, which resulted in excellent predictions (R-2 > 0.99). The prediction parameters for the sample sets were: the RMSEC were in the range 0.59-2.09% and, RMSECV were in the range 0.92-5.60%, hence on the basis of the PLSR models, quantification of adulterant could be determined at levels < 0.59%.Scientific Research Project Center of Karamanoglu Mehmetbey University [18-M-18]; Scientific Research Project Center of Ankara University; TUBITAKTurkiye Bilimsel ve Teknolojik Arastirma Kurumu (TUBITAK) [TUBITAK-116Z159]; TUBITAK under the 2219-Research Fellowship ProgramThis study is supported financially by the Scientific Research Project Center of Karamanoglu Mehmetbey University (Project number 18-M-18). The authors would like to thank the Scientific Research Project Center of Ankara University and TUBITAK (Project number TUBITAK-116Z159) for providing the financial support to gain Unscrambler (R) X10.4 (CAMO software, Oslo, Norway) program. The authors would also like to thank TUBITAK under the 2219-Research Fellowship Program for providing the financial support to carry out this research work