Compositional analysis in sorghum (Sorghum bicolor) NIR spectral techniques based on mean spectra from single seeds

Abstract

Sorghum (Sorghum bicolor) is an economically important cereal crop that can be used as human food, animal feed, and for industrial use such as bioenergy. In sorghum breeding programs, development of cultivars with desirable seed quality characteristics is important and development of rapid low-cost screening methods for seed nutritional traits are desired, since most standard methods are destructive, slow, and less environmentally friendly. This study investigates the feasibility of single kernel NIR spectroscopy (SKNIRS) for rapid determination of individual sorghum seed components. We developed successful multivariate prediction models based on partial least squares (PLS) regression for protein, oil, and weight in sorghum. The results showed that for sorghum protein content ranging from 8.92% to 18.7%, the model coefficients of determination obtained were $R_{CAL}^2=0.95$ (RMSEC= 0.44) and $R_{PRED}^2=0.87$ (RMSEP= 0.69). The model coefficients of determination for oil prediction were $R_{CAL}^2=0.92$ (RMSEC= 0.23) and $R_{PRED}^2=0.71$ (RMSEP= 0.41) for oil content ranging from 1.96% to 5.61%. For weight model coefficients of determination were $R_{CAL}^2=0.81$ (RMSEC= 0.007) and $R_{PRED}^2=0.63$ (RMSEP= 0.007) for seeds ranging from 4.40 mg to 77.0 mg. In conclusion, mean spectra SKNIRS can be used to rapidly determine protein, oil, and weight in intact single seeds of sorghum seeds and can provide a nondestructive and quick method for screening sorghum samples for these traits for sorghum breeding and industry use.

Keywords: near infrared spectroscopy; oil; phenotyping; seed content; sorghum protein; sustainable food systems.