Use of near-infrared spectroscopy for the rapid evaluation of soybean [Glycine max (L.) Merri.] water soluble protein content.

Water soluble protein content (WSPC) is a parameter of great significance to the soybean food industry. So far, genetic studies and breeding practices have been limited by the lack of a rapid technique for the evaluation of WSPC. Near-infrared reflectance spectroscopy (NIRS) is widely applied for rapid quantification of many traits, including moisture, protein and oil content, and dietary fiber. The present study aimed to establish and evaluate a NIRS regression model for the rapid prediction of WSPC in soybean. Results showed that seed coat color had a profound impact on the accuracy of protein content prediction, whereas the seed coat itself deeply influenced protein determination. We established a partial least squares (PLS) regression model with 167 soybean samples whose seed coat had been removed. Based on multiplicative scatter correction and Savitsky-Golay transformation, the highest determination coefficient (R2) was 0.831, and the relative predictive determinant was 2.417. Further analysis showed that seed roundness correlated negatively with WSPC (r=-0.59, P<0.001) and greatly impacted PLS regression model prediction accuracy. The PLS model was suitable only for intact seeds whose coat had been peeled off, but not for broken seeds, soy powder, and green cotyledon soybean seeds. This study highlights the effect the seed coat has on soybean composition determination by NIRS. Moreover, the established PLS model for soybean WSPC determination could facilitate genetic studies and breeding.