High-Throughput Near-Infrared Reflectance Spectroscopy for Predicting Quantitative and Qualitative Composition Phenotypes of Individual Maize Kernels

Near-infrared reflectance (NIR) spectroscopy can be used for fast and reliable prediction of organic compounds in complex biological samples. We used a recently developed NIR spectroscopy instrument to predict starch, protein, oil, and weight of individual maize (Zea mays) seeds. The starch, protein, and oil calibrations have reliability equal or better to bulk grain NIR analyzers. We also show that the instrument can differentiate quantitative and qualitative seed composition mutants from normal siblings without a specific calibration for the constituent affected. The analyzer does not require a specific kernel orientation to predict composition or to differentiate mutants. The instrument collects a seed weight and a spectrum in 4-6 see and can collect NIR data alone at a 20-fold faster rate. The spectra are acquired while the kernel falls through a glass tube illuminated with broad spectrum light. These results show significant improvements over prior single-kernel NIR systems, making this instrument a practical tool to collect quantitative seed phenotypes at high throughput. This technology has multiple applications for studying the genetic and physiological influences on seed traits.