Development of calibration equations to predict oil content and fatty acidcomposition in Brassicaceae germplasm by near-infrared reflectance spectroscopy

The use of nondestructive analytical methods is critical for the evaluation of very small seed samples such as those from germplasm collections. The o bjective of this study was to evaluate the potential of near-infrared refle ctance spectroscopy (NIRS) for the simultaneous analysis of seed oil conten t and concentration of major fatty acids in intact-seed samples of the fami ly Brassicaceae. A total of 495 samples from 56 genera and 128 species were analyzed by NIRS. The fatty acid composition of the seed oil was determine d in all the samples by gas-liquid chromatography (GLC). The total seed oil content was determined by solvent extraction in 129 samples from 22 genera . Calibration equations for oil content (n = 97) and individual fatty acids (n = 410) were developed and tested through external validation with the s amples not included in the calibration sets. The calibration equations for oil content (r(2) = 0.97 in validation) and concentrations of C-18:1 (r(2) = 0.93), C-18:3 (r(2) = 0.95), and C-22:1 (r(2) = 0.94) showed very good pe rformance and provided reliable estimations of these traits in the samples of the validation set. The calibration equations for C-16:0, C-18:0, and C- 18:2 content were less reliable, with r(2) from 0.67 to 0.73. There was pra ctically no response of NIRS to differences in C-20:1 (r(2) = 0.31). These results demonstrated that the oil content and concentrations of C-18:1, C-1 8:3, and C-22:1 can be estimated reliably within the family Brassicaceae by using NIRS calibration equations integrating broad taxonomic variability.