Near-infrared reflectance spectroscopy calibrations for assessment of oil, phenols, glucosinolates and fatty acid content in the intact seeds of oilseed Brassica species

Using near-infrared reflectance spectroscopy (NIRS) to predict the nitrogen levels in the stem and root tissues of Brassica juncea (Indian mustard)

Brassica juncea depends heavily on nitrogen (N) fertilizers for growth and accumulation of seed protein. However, it is an inefficient mobilizer of applied N which leads to accumulation of excess N in the soil, posing environmental risks. Hence, it is imperative to systematically examine spatial-temporal pattern of crop N to efficiently manage N application. The Kjeldahl method is commonly used to estimate N status of crops but it is a destructive method that entails the use of perilous and expensive chemicals. Near-infrared reflectance spectroscopy (NIRS) offers a safe, accurate, and non-destructive alternative for large-scale screening of seed metabolites. Currently, no NIRS model exists to quickly estimate N content in shoots and roots from large germplasm sets in any rapeseed-mustard crop. Developing such a model is essential to breed for enhanced nitrogen use efficiency (NUE). We used 738 shoot and 346 root samples from a B. juncea diversity set to construct the NIRS models. A diverse range of genetic variation in N content was recorded in the stem (0.21-6.61%) and root (0.15-3.04%) tissues of the crop raised on two different N levels (N0 and N100). Modified partial least squares (MPLS) method was employed to establish a regression equation linking reference N values with spectral changes. The developed models exhibited strong associations with reference values, with RSQ values of 0.884 for stem and 0.645 for roots. Furthermore, external validation confirms the reliability of the developed models. The developed models have strong predictive capabilities for rapid and reliable N estimation in various tissues of B. juncea plants.

Determination of glucosinolates and isothiocyanates in glucosinolate-rich vegetables and oilseeds using infrared spectroscopy: A systematic review

Cruciferous vegetables and oilseeds are rich in glucosinolates that can transform into isothiocyanates upon enzymic hydrolysis during post-harvest handling, food preparation and/or digestion. Vegetables contain glucosinolates that have beneficial bioactivities, while glucosinolates in oilseeds might have anti-nutritional properties. It is therefore important to monitor and assess glucosinolates and isothiocyanates content through the food value chain as well as for optimized crop production. Vibrational spectroscopy methods, such as infrared (IR) spectroscopy, are used as a nondestructive, rapid and low-cost alternative to the current and common costly, destructive, and time-consuming techniques. This systematic review discusses and evaluates the recent literature available on the use of IR spectroscopy to determine glucosinolates and isothiocyanates in vegetables and oilseeds. NIR spectroscopy was used to predict glucosinolates in broccoli, kale, rocket, cabbage, Brussels sprouts, brown mustard, rapeseed, pennycress, and a combination of Brassicaceae family seeds. Only one study reported the use of NIR spectroscopy to predict broccoli isothiocyanates. The major limitations of these studies were the absence of the critical evaluation of errors associated with the reference method used to develop the calibration models and the lack of interpretation of loadings or regression coefficients used to predict glucosinolates.

A Review of Machine Learning for Near-Infrared Spectroscopy

The analysis of infrared spectroscopy of substances is a non-invasive measurement technique that can be used in analytics. Although the main objective of this study is to provide a review of machine learning (ML) algorithms that have been reported for analyzing near-infrared (NIR) spectroscopy from traditional machine learning methods to deep network architectures, we also provide different NIR measurement modes, instruments, signal preprocessing methods, etc. Firstly, four different measurement modes available in NIR are reviewed, different types of NIR instruments are compared, and a summary of NIR data analysis methods is provided. Secondly, the public NIR spectroscopy datasets are briefly discussed, with links provided. Thirdly, the widely used data preprocessing and feature selection algorithms that have been reported for NIR spectroscopy are presented. Then, the majority of the traditional machine learning methods and deep network architectures that are commonly employed are covered. Finally, we conclude that developing the integration of a variety of machine learning algorithms in an efficient and lightweight manner is a significant future research direction.

Prediction of maize flour adulteration in chickpea flour (besan) using near infrared spectroscopy

The present study was performed to develop Near-infrared spectroscopy based prediction method for the quantification of the maize flour adulteration in chickpea flour. Adulterated samples of Chickpea flour (besan) were prepared by spiking different concentrations of maize flour with pure Chickpea flour in the range of 1–90% (w/w). The spectra of pure Chickpea flour, pure maize flour, and adulterated samples of Chickpea flour with maize flour were acquired as the logarithm of reciprocal of reflectance (log 1/R) in the entire Visible-NIR wavelength range of 400–2498 nm. The acquired spectra were pre-processed by Ist derivative, standard normal variate, and detrending. The calibration models were developed using modified partial least square regression (MPLSR), partial least square regression and principal component regression. The optimal model was selected on the basis of highest values of the coefficient of determination (RSQ), one minus variance ratio (1-VR) and lowest values of standard errors of calibration (SEC), and standard error of cross-validation (SECV). MPLSR model having RSQ and 1-VR value of 0.999 and 0.996 having SEC and SECV value of 1.092 and 2.042 was developed for quantification of maize flour adulteration in chickpea flour. Cross validation and external validation of the developed models resulted in RSQ of 0.999, 0.997 and standard error of prediction of 1.117, and 2.075, respectively.

Genomic Regions Associated With Seed Meal Quality Traits in Brassica napus Germplasm

The defatted Brassica napus (rapeseed) meal can be high-protein feed for livestock as the protein value of rapeseed meal is higher than that of the majority of other vegetable proteins. Extensive work has already been carried out on developing canola rapeseed where the focus was on reducing erucic acid and glucosinolate content, with less consideration to other antinutritional factors such as tannin, phytate, sinapine, crude fiber, etc. The presence of these antinutrients limits the use and marketing of rapeseed meals and a significant amount of it goes unused and ends up as waste. We investigated the genetic architecture of crude protein, methionine, tryptophan, total phenols, β-carotene, glucosinolates (GLSs), phytate, tannins, sinapine, and crude fiber content of defatted seed meal samples by conducting a genome-wide association study (GWAS), using a diversity panel comprising 96 B. napus genotypes. Genotyping by sequencing was used to identify 77,889 SNPs, spread over 19 chromosomes. Genetic diversity and phenotypic variations were generally high for the studied traits. A total of eleven genotypes were identified which showed high-quality protein, high antioxidants, and lower amount of antinutrients. A significant negative correlation between protein and limiting amino acids and a significant positive correlation between GLS and phytic acid were observed. General and mixed linear models were used to estimate the association between the SNP markers and the seed quality traits and quantile-quantile (QQ) plots were generated to allow the best-fit algorithm. Annotation of genomic regions around associated SNPs helped to predict various trait-related candidates such as ASP2 and EMB1027 (amino acid biosynthesis); HEMA2, GLU1, and PGM (tryptophan biosynthesis); MS3, CYSD1, and MTO1 (methionine biosynthesis); LYC (β-carotene biosynthesis); HDR and ISPF (MEP pathway); COS1 (riboflavin synthesis); UGT (phenolics biosynthesis); NAC073 (cellulose and hemicellulose biosynthesis); CYT1 (cellulose biosynthesis); BGLU45 and BGLU46 (lignin biosynthesis); SOT12 and UGT88A1 (flavonoid pathway); and CYP79A2, DIN2, and GSTT2 (GLS metabolism), etc. The functional validation of these candidate genes could confirm key seed meal quality genes for germplasm enhancement programs directed at improving protein quality and reducing the antinutritional components in B. napus.

A comprehensive Vis-NIRS equation for rapid quantification of seed glucosinolate content and composition across diverse Brassica oilseed chemotypes

The globally cultivated Brassica crops contain high deliverable concentrations of health-promoting glucosinolates. Development of a Visible-Near InfraRed Spectroscopy (Vis-NIRS) calibration to profile different glucosinolate components from 641 diverse Brassica juncea chemotypes was attempted in this study. Principal component analysis of HPLC-determined glucosinolates established the distinctiveness of four B. juncea populations used. Subsequently, modified partial least square regression based population-specific and combined Vis-NIRS models were developed, wherein the combined model exhibited higher coefficient of determination (R²; 0.81-0.97) for eight glucosinolates and higher ratio of prediction determination (RPD; 2.42-5.35) for seven glucosinolates in B. juncea populations. Furthermore, range error ratio (RER > 4) for twelve and RER > 10 for eight glucosinolates make the combined model acceptable for screening and quality control. The model also provided excellent prediction for aliphatic glucosinolates in four oilseed Brassica species. Overall, our work highlights the potential of Vis-NIR spectroscopy in estimating glucosinolate content in the economically important Brassica oilseeds.

Development and Validation of Kompetitive Allele-Specific PCR Assays for Erucic Acid Content in Indian Mustard [Brassica juncea (L.) Czern and Coss.]

Brassica juncea L. is the most widely cultivated oilseed crop in Indian subcontinent. Its seeds contain oil with very high concentration of erucic acid (≈50%). Of late, there is increasing emphasis on the development of low erucic acid varieties because of reported association of the consumption of high erucic acid oil with cardiac lipidosis. Erucic acid is synthesized from oleic acid by an elongation process involving two cycles of four sequential steps. Of which, the first step is catalyzed by β-ketoacyl-CoA synthase (KCS) encoded by the fatty acid elongase 1 (FAE1) gene in Brassica. Mutations in the coding region of the FAE1 lead to the loss of KCS activity and consequently a drastic reduction of erucic acid in the seeds. Molecular markers have been developed on the basis of variation available in the coding or promoter region(s) of the FAE1. However, majority of these markers are not breeder friendly and are rarely used in the breeding programs. Present studies were planned to develop robust kompetitive allele-specific PCR (KASPar) assays with high throughput and economics of scale. We first cloned and sequenced FAE1.1 and FAE1.2 from high and low erucic acid (<2%) genotypes of B. juncea (AABB) and its progenitor species, B. rapa (AA) and B. nigra (BB). Sequence comparisons of FAE1.1 and FAE1.2 genes for low and high erucic acid genotypes revealed single nucleotide polymorphisms (SNPs) at 8 and 3 positions. Of these, three SNPs for FAE1.1 and one SNPs for FAE1.2 produced missense mutations, leading to amino acid modifications and inactivation of KCS enzyme. We used SNPs at positions 735 and 1,476 for genes FAE1.1 and FAE1.2, respectively, to develop KASPar assays. These markers were validated on a collection of diverse genotypes and a segregating backcross progeny. KASPar assays developed in this study will be useful for marker-assisted breeding, as these can track recessive alleles in their heterozygous state with high reproducibility.

Association Mapping of Seed Quality Traits Under Varying Conditions of Nitrogen Application in Brassica juncea L. Czern & Coss

Indian mustard (Brassica juncea) is a major source of vegetable oil in the Indian subcontinent. The seed cake left after the oil extraction is used as livestock feed. We examined the genetic architecture of oil, protein, and glucosinolates by conducting a genome-wide association study (GWAS), using an association panel comprising 92 diverse genotypes. We conducted trait phenotyping over 2 years at two levels of nitrogen (N) application. Genotyping by sequencing was used to identify 66,835 loci, covering 18 chromosomes. Genetic diversity and phenotypic variations were high for the studied traits. Trait performances were stable when averaged over years and N levels. However, individual performances differed. General and mixed linear models were used to estimate the association between the SNP markers and the seed quality traits. Population structure, principal components (PCs) analysis, and discriminant analysis of principal components (DAPCs) were included as covariates to overcome the bias due to the population stratification. We identified 16, 23, and 27 loci associated with oil, protein, and glucosinolates, respectively. We also established LD patterns and haplotype structures for the candidate genes. The average block sizes were larger on A-genome chromosomes as compared to the B- genome chromosomes. Genetic associations differed over N levels. However, meta-analysis of GWAS datasets not only improved the power to recognize associations but also helped to identify common SNPs for oil and protein contents. Annotation of the genomic region around the identified SNPs led to the prediction of 21 orthologs of the functional candidate genes related to the biosynthesis of oil, protein, and glucosinolates. Notable among these are: LACS5 (A09), FAD6 (B05), ASN1 (A06), GTR2 (A06), CYP81G1 (B06), and MYB44 (B06). The identified loci will be very useful for marker-aided breeding for seed quality modifications in B. juncea.

Targeted Knockout of BnTT2 Homologues for Yellow-Seeded Brassica napus with Reduced Flavonoids and Improved Fatty Acid Composition

Brassica napus is one of the important oil crops grown worldwide, and oil quality improvement is a major goal in rapeseed breeding. Yellow seed is an excellent trait, which has great potential in improving seed quality and economic value. In this study, we created stable yellow seed mutants using a CRISPR/Cas9 system and obtained the yellow seed phenotype only when the four alleles of two BnTT2 homologues were knocked out, indicating that the two BnTT2 homologues had conserved but redundant functions in regulating seed color. Histochemical staining and flavonoid metabolic analysis proved that the BnTT2 mutation hindered the synthesis and accumulation of proanthocyanidins. Transcriptome analysis also showed that the BnTT2 mutation inhibited the expression of genes in the phenylpropanoid and flavonoid biosynthetic pathway, which might be regulated by the complex of BnTT2, BnTT8 and BnTTG1. In addition, the homozygous mutants of BnTT2 homologues increased oil content and improved fatty acid composition with higher linoleic acid (C18:2) and linolenic acid (C18:3), which could be used for the genetic improvement of rapeseed. Overall, this research showed that the BnTT2 mutation can be used for yellow seed breeding and oil improvement, which is of great significance in improving the economic value of rapeseeds.

Rapid Determination of the Oil and Moisture Contents in Camellia gauchowensis Chang and Camellia semiserrata Chi Seeds Kernels by Near-infrared Reflectance Spectroscopy

A fast and effective determination method of different species of vegetable seeds oil is vital in the plant oil industry. The near-infrared reflectance spectroscopy (NIRS) method was developed in this study to analyze the oil and moisture contents of Camelliagauchowensis Chang and C.semiserrata Chi seeds kernels. Calibration and validation models were established using principal component analysis (PCA) and partial least squares (PLS) regression methods. In the prediction models of NIRS, the levels of accuracy obtained were sufficient for C.gauchowensis Chang and C.semiserrata Chi, the correlation coefficients of which for oil were 0.98 and 0.95, respectively, and those for moisture were 0.92 and 0.89, respectively. The near infrared spectrum of crush seeds kernels was more precise compared to intact kernels. Based on the calibration models of the two Camellia species, the NIRS predictive oil contents of C.gauchowensis Chang and C.semiserrata Chi seeds kernels were 48.71 ± 8.94% and 58.37 ± 7.39%, and the NIRS predictive moisture contents were 4.39 ± 1.08% and 3.49 ± 0.71%, respectively. The NIRS technique could determine successfully the oil and moisture contents of C.gauchowensis Chang and C.semiserrata Chi seeds kernels.