Compositional analyses reveal relationships among components of blue maize grains

Authors: NANKAR, AMOL - Institute Of Plant Biology And Biotechnology; Scott, Marvin; PRATT, RICHARD - New Mexico State University

Submitted to: Plants
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/11/2020
Publication Date: 12/14/2020
Citation: Nankar, A.N., Scott, M.P., Pratt, R.C. 2020. Compositional analyses reveal relationships among components of blue maize grains. Plants. 9(12). Article 1775. https://doi.org/10.3390/plants9121775.
DOI: https://doi.org/10.3390/plants9121775

Interpretive Summary: The chemical composition of grain determines how useful it is for different purposes. Near infrared spectroscopy is frequently used to analyze the composition of grain, however it requires a mathematical equation called a calibration in order to give useful information. The goal of this work was to develop calibration models to allow prediction of valuable chemical components in pigmented maize kernels. We tested different methods of calibration development and learned which method performed best for each chemical component examined. This allowed us to examine the relationship among grain components in the varieties studied. The new calibration models and information about the relationship among components provide new tools for studying grain composition in pigmented maize. These tools will be useful to plant breeders and scientists interested in developing maize varieties with improved nutritional quality and functional properties.

Technical Abstract: The aim of this experiment was to develop NIRS (near infrared spectroscopy) calibrations for proximates in pigmented maize. A total of 143 samples from eight blue, red, and purple maize accessions were evaluated in four locations across New Mexico during 2013 and 2014 to develop calibration equations to evaluate 20 compositional constituents including starch, protein, oil, total fatty acids, essential and conditionally essential amino acids. Based on reference analysis, different estimation models were developed using multiple linear regression (MLR), principal component regression (PCR), and partial least squares (PLS). In all models, the highest accuracy was attained for glycine, cysteine, methionine, total fatty acids and oil content for MLR (r=0.91 and 0.90), glycine and methionine for PCR (r=0.51) and glycine for PLS (r=0.59). For protein, starch and anthocyanin, MLR models yielded the best results with rMLR of 0.87, 0.86 and 0.86, respectively. Performance of calibration models were determined based on coefficient of determination (R2), root mean square of calibration set (RMSEC), root mean square of validation set (RMSEP), and ratio of standard deviation to the RMSEP (RPD). The R2 of the prediction equations for MLR model ranged from 0.47 for leucine to 0.92 for methionine whereas the R2 for PCR models ranged from 0.04 for proline to 0.51 for methionine, and the R2 for PLS ranged from 0.16 for starch to 0.60 for methionine. Across all calibration models, methionine showed the highest R2 among all constituents. Based on R2 and RMSEC, total amino acids, fatty acids, protein, oil, starch and anthocyanin appear to have reliable blue maize predictions. Multivariate analysis was also used to determine the divergence in kernel compositional traits among evaluated populations. Four principal components with an eigenvalue > 1 were identified, which explained 85.26% variation to the total cumulative variance with the first two components contributing 49.62% and 22.20% variance, respectively. The relationship between kernel compositional traits was explained by correlation networks and highly correlated traits were identified. Considering non-destructive nature, rapid screening and the ability to analyze multiple traits concurrently, NIRS is ideal for breeders interested in grain composition.