Protein Content of Wheat by Near-Infrared Spectroscopy of Whole Grain: Collaborative Study

A collaborative study was performed to assess accuracy, repeatability, and reproducibility of a nearinfrared (near-IR) method for determining crude protein content (PC) of whole-grain wheat. Four types of commercially available near-IR instruments, representing various combinations of wavelength region, mode of energy capture, method of energy dispersion, and treatment of spectral data, were used. Eight, 9,10, and 11 collaborators were involved, the exact number depending on instrument type. All collaborators received 22 samples of whole-grain hard red winter (HRW) wheat. They were furnished reference PCs (i.e., protein concentrations, w/w) corrected to a 12% moisture basis for instrument standardization. AOAC Method 990.03— combustion analysis—was the reference procedure. Standardization consisted of performing one of the following treatments to the instrument manufacturer's (or federal agency's) PC equation: (1) bias correction, (2) slope and intercept correction, or (3) recalibration with inclusion of standardization sample spectra. Standardized equations were then applied to a test set of 12 unknown HRW wheat sample spectra, with 2 samples blindly duplicated. The PCs of test samples ranged from 9 to 16%. Near-IR predictions were compared with reference measurements. Averaged within instrument type, root mean square of differences were 0.22, 0.24, 0.25, and 0.26% PC, depending on instrument. Corrected for bias within the test set, standard errors became 0.22, 0.18, 0.21, and 0.24% PC, respectively. These values were approximately twice the estimated lower limit for error (representing sample inhomogeneity). Overall repeatability relative standard deviation (RSD,) values were 0.92, 0.36, 0.42, and 0.74%, respectively. Overall reproducibility relative standard deviation (RSDR) values were 1.15, 0.61,1.53, and 1.38%. Such values for within-laboratory and between-laboratory variations of the near- IR methods were equivalent to values reported for the combustion method (990.03) for wheat. An inhouse study that examined all 6 U.S. wheat classes with one of the 4 instrument types produced repeatability and reproducibility values similar to those of the collaborative study, suggesting that the near-IR technique may be applied to red, white, hard, soft, and durum wheats. The near-IR method for determination of PC of whole-grain wheat has been adopted First Action (997.06) by AOAC INTERNATIONAL.