This project tested NIR performance of two commonly utilized calibrations for protein and amino acid estimation in whole soybeans. This project tested performance of three presentation methods or grinds for the PerkinElmer calibration (whole seed, plate mill, and centrifugal mill) while the Evonik calibration was tested only with soybean seed ground with the centrifugal mill.
The primary focus of this work was to determine whether the Evonik calibration that is provided to feed mills to estimate amino acid levels in soybeans and soybean meal for ration development (including additions of supplemental synthetic amino acids) was biased toward an underestimation of amino acids within the soybean. We found the calibration to perform with a high level of accuracy. The Evonik calibration slightly underestimated protein, threonine, and tryptophan while slightly overestimating the other primary amino acids. Precision between lab and NIR was estimated here with the Coefficient of Determination or R2. The Evonik calibration tended to perform with a high level of precision as well. While this differed between individual amino acids. Those that are more difficult to assess with traditional lab methods such as methionine and cysteine were not predicted effectively with this calibration, but others were predicted well.
The PerkinElmer calibration performed similarly with the Evonik, but in most cases biases were slightly lager and R2 values slightly lower. Despite vast differences in NIR hardware, calibration development and reference labs, the two instrument/calibration combinations tended to perform similar to one another.