Effect of Sample Type and Chemometric Method in Determination of Proteinand Oil Ratio in Maize using FT-NIR (Fourier Transform-Near Infrared) Spectroscopy

Ft-nir (fourier transform-near infrared) spectroscopy is an actively used instrument in detection of sample ingredients in food and industrial areas. in this study, 260 maize samples including hybrids and inbreds with different genetic characteristics were used as material. on these samples, quantitative protein and oil analyses were carried out. calibration models (n=227) were developed using these results and spectral data collected with ft-nir instrument from seed and flour samples. multiple linear regression (mlr) and partial least squares regression (plsr) methods were utilized for the purpose of developing models. reliability of the models was tested with external validation procedure (n=20). results showed that mlr (rmsec=0.5482; see=0.5494; r=0.882; r^2=0.7776) and plsr (rmsec=0.5504; see=0.5516; r=0.880; r^2=0.7758) methods were similar in estimation power for protein ratio when the models were developed from flour samples, while for oil ratio plsr model (rmsec=0.4429; see=0.4439; r=0.719; r^2=0.5179) gave a higher power with respect to mlr method. after external validation, there was no change in model comparison for protein ratio, whereas ml model emerged to be more reliable for oil ratio analysis. the models based on the spectra taken from intact seed samples were not suitable for use. we concluded that statistical method and sample type had significant effects on estimation power, and the measurement method used here yielded more reliable results on flour samples.