Near Infrared Reflectance Spectroscopy of Oil in Intact Canola Seed (Brassica Napus L.). II. Association between Principal Components and oil Content

Abstract

This work reports that the measurements of the likeness or the uniqueness of the 1100–2500 nm reflectance spectra of intact canola seed determined from principal component analysis (PCA) approximated spectra with global H or neighbourhood H statistics were not associated with oil concentration within the seed. The absence of stability in association between the H measurements and oil content was related to inconsistency in the amount and distribution (between principal components) of the spectral variation correlated to the oil content within and between different batches of canola seed. PCA was used to approximate variation in the 1100–2500 nm, second order derivative, reflectance spectra of intact canola seed, acquired from 15 batches of seed samples. The first eight principal components (PCs) captured 97.14% to 99.35% of the total variance in the spectra. The amount of variation captured by individual components was independent of the number of samples in the batch and oil content within the seed. The pattern of variance distribution among principal components was inconsistent and highlighted the uniqueness of the origin of the spectral variation in each batch of canola seed. In this study, the strength of correlation between oil content and principal components was used as a measure of component significance to the analysis of oil in the intact canola seed. In the examined sets of spectra, oil content was correlated to the low order components, PC1 to PC4. In the 15 files of spectra, oil content showed the strongest correlation to PC2 in eight sets of data, to PC3 in four sets of data and to PC1 in three sets of data. The strength of association between oil and the individual components varied considerably in magnitude among examined files of spectra; r2 = 0.28–0.81 for the first strongly correlated component, r2 = 0.05–0.29 for the second and r2 = 0.02–0.19 for the third. The position of the PCs in the correlation sequence was inconsistent and underlined differences of oil signal/spectral data interactions in the individual sets of data. Examination of principal component loadings showed that in the reported files of spectra, principal components correlated to the oil content frequently captured variance at segments, which denote absorptions specific and accidental to canola oil. The outline of the loadings did not conform to a single, regular pattern common to all sets of data. The reported results are in disagreement with rationale of the methodology, which involves the spectra matching techniques for validating the predictive efficiency of near infrared (NIR) calibrations. The reported results highlighted that the reliable NIR quantification of oil content from reflectance spectra of intact canola seed would require an independent validation for every acquired set of spectra.

Development and Evaluation of Dedicated Bench-Type near Infrared Spectrometers to Estimate the Flavour of Green Soybeans from Single Pod Measurements

“…Sato et al 3 examined the feasibility of NIR spectroscopy for the determination of fatty acid (FA)� composition of sesame seed and reported that it could be estimated with reasonable accuracy for breeding. Golebiowski et al 4,5 investigated oil in intact canola seed using NIR spectroscopy and discussed the attributes of the spectrum and the association between principal components and oil content. Arganosa et al 6 found that the accuracy of NIR re��ectance spectroscopy for the determination of crude protein content of field peas was adequate for breeding programmes.…”

Genomic Prediction and Genetic Correlation of Agronomic, Blackleg Disease, and Seed Quality Traits in Canola (Brassica napus L.)

“…Seed quality traits were measured at MI16, HrI16, HrI17 and Hr17 and determined by near infrared spectroscopy on whole seeds using the Foss XDS Rapid Content Analyser (Foss Denmark, Hillerod, Denmark) according to Golebiowski et al, 2005 [ 12 ]. The algorithms for each seed trait were developed using the peer reviewed protocols as described for each trait.…”

“…The approach uses a reference population, which is phenotyped and genotyped, to predict the genetic values of selection candidates which are only genotyped. Recent studies in canola [12] demonstrated that GS accuracy is in the usable range for reducing the breeding cycle [13][14][15]. The role of GS in accelerating crop improvement, the factors affecting prediction accuracy, and possible strategies for future application have been reviewed [16][17][18].…”

In Situ Estimation of the Constituents of Green Soybean (Edamame) Pod using Near-Infrared Transmission Spectroscopy

“…Sato et al examined the feasibility of NIR spectroscopy for estimating sesame seed fatty acid (FA) composition, and reported that it was reasonably accurate for breeding purposes (Sato et al, 2003). Golebiowski (2004) and Golebiowski et al (2005) used NIR spectroscopy to investigate the oil in intact canola seed and discussed the attributes of the spectrum and the relationship between the principal components and the oil content. Arganosa et al (2006) found that NIR reflectance spectroscopy was sufficiently accurate for determining the crude protein content of field peas for a breeding program.…”