ABSTRACT:

Near infrared reflectance (NIR) spectroscopy was used for the analysis of fatty acids in whole seed Brassica napus and Brassica juncea cultivars. Calibrations were developed for the following fatty acids: palmitic (C16:0), stearic (C18:0), oleic (C18:1), linoleic (C18:2), linolenic (C18:3), arachidic (C20:0), eicosenoic (C20:1) and erucic (C22:1). The spectra of cultivars of different botanical origin and quality type were heterogeneous; therefore calibrations were developed using full spectral information and modified partial least squares (mPLS) for each separate group of Brassicas. The calibrations could be applied to other spectrally similar populations and their accuracy allowed the rapid screening of approximately 80,000 samples of Brassica seed, produced each year by the oilseed breeding programs situated in Horsham.

INTRODUCTION

Brassica oilseed crops are primarily grown for the production of two distinct types of vegetable oils, which are used either for human consumption or for industrial applications. Edible Brassica oils contain a high level of oleic acid (55 - 72 %), while eicosenoic (4 - 9 %) and erucic (45 - 55 %) acids dominate the industrial type Brassica oils. The standard procedure for the analysis of fatty acids in Brassica oils is gas chromatography.

Near infrared reflectance (NIR) spectroscopy, has been successfully applied as an alternative technique to gas chromatography for the analysis of fatty acids in a number of oilseed crops. The main species studied to date are canola (Brassica napus; Panford and deMan, 1990; Reinhardt and Röbbelen, 1991; Golebiowski et al, 1995), sunflower (Helianthus annuus; Panford and deMan, 1990), soybean (Glycine max.; Sato et al, 1991) and Ethiopian mustard (Brassica carinata; Velasco et al, 1995).

Selection of specific wavelengths is the most common approach to date for calibration development. While these calibrations were reported to be relatively accurate, there is no information supplied that this method was used for routine analysis of fatty acids in oilseed crops (Sato et al, 1991, Velasco et al, 1995). Alternatively, it has been reported that full spectrum information can successfully produce accurate calibrations for the routine analysis of fatty acids in canola. Use of the full spectrum had the advantage of allowing better management of spectral information and determination of unusual spectral properties (Golebiowski et al, 1995).

This study is concerned with the problem of whether NIR full spectrum analysis can be applied to fatty acid determination in high erucic acid rapeseed and canola (Brassica napus cultivars) and also condiment and canola quality mustard (Brassica juncea cultivars).

MATERIALS AND METHODS

Canola, high erucic rapeseed, canola quality and condiment mustard samples (totalling approximately 80,000) were collected from throughout the southern part of Australia and some from Canada.

All seed samples were scanned on a Model 5000 monochromator (NIRSystems Inc., Silver Spring, MD, USA) equipped with a semi-automatic spinning 50 mm ring cup module. The unit operated in reflectance mode and a wavelength range between 1100 and 2500 nm. Supporting spectrometric and chemometric software (Infrasoft International ver. 3.1, USA) was used to treat the spectra with the CENTRE algorithm describing the spectral boundaries and detecting outliers. Representative samples were identified entirely from spectral information by the SELECT algorithm (Shenk and Westerhaus, 1995).

The eight major individual fatty acids (C16:0, C18:0, C18:1, C18:2, C18:3, C20:0, C20:1 and C22:1) of selected samples were analysed as fatty acid methyl esters (FAME; Daun and Mazur, 1983) on a Model 14A gas chromatograph (GC). This unit was equipped with a Model AOC-20s autosampler, FID detector (Shimadzu, Kyoto, Japan) and fitted with a capillary column; BPX70: 70 per cent cyanopropylpolysiloxane; film thickness 0.25 μm; I.D. 0.32 mm; length 25 m (Scientific Glass Engineering, Ringwood, Australia). Data was collected and processed with a Model C-R4A computing integrator (Shimadzu, Kyoto, Japan). The GC data were matched with spectral information and equations were developed using the mPLS algorithm and cross validation technique as described by Golebiowski et al (1995).

RESULTS AND DISCUSSION

The four different types of brassica crops (canola, high erucic acid rapeseed, canola quality juncea and condiment juncea), produced distinctly different NIR spectral patterns and thus each of them required separate calibrations for fatty acid analysis.

The results of the calibration for the high oleic acid canola and canola quality mustard are shown in Table 1. The major fatty acids; oleic acid (C18:1, highlighted as it is considered an important monounsaturated dietary fat in these two brassica cultivars), linoleic acid (C18:2) and linolenic acid (C18:3), displayed high agreement between their predicted values using the calibration and the actual values obtained by GC. This was demonstrated by the relatively low standard error of calibration (SEC, the ideal value should be approaching zero) and high RSQ values (ideal value should be approaching 1). The remaining fatty acids (C16:0, C18:0, C20:0 and C20:1) were considered to be minor constituents due to their low concentration in the seed. While these tended to have low RSQ values, due to restricted variance, they also exhibited low SEC values, which resulted in predictions that were quite accurate.

The results of the calibration for the high erucic acid rapeseed and condiment mustard were shown in Table 2. For these cultivars, the major fatty acids were oleic acid (C18:1), linoleic acid (C18:2), linolenic acid (C18:3) and erucic acid (C22:1). These also displayed good agreement between their predicted values using the calibration and the actual values obtained by GC.

In the high erucic acid rapeseed, C22:1 is an important fatty acid as a basic raw material in the production of polymers and paints and is not considered a dietary fatty acid. Condiment juncea tends to have high levels of C22:1 but is usually consumed in such small quantities that little erucic acid is ingested.

Table 1. Calibration statistics for fatty acids in high oleic acid canola and canola quality

juncea

N = number of samples; Mean = average of fatty acid percentage; SEC = standard error of calibration; RSQ = coefficient of determination of calibration; SECV = standard error of cross validation; 1-VR = coefficient of determination of cross validation.

Figure 1: Comparison between gas chromatography data and NIR predicted values of oleic acid (C18:1) for the high oleic acid canola.

Figure 2: Comparison between gas chromatography data and NIR predicted values of oleic acid (C18:1) for the high oleic acid canola quality mustard.

Table 2. Calibration statistics for fatty acids in high erucic acid rapeseed and condiment mustard

N = number of samples; Mean = average of fatty acid percentage; SEC = standard error of calibration; RSQ = coefficient of determination of calibration; SECV = standard error of cross validation; 1-VR = coefficient of determination of cross validation.

Figure 3: Comparison between gas chromatography data and NIR predicted values of erucic acid (C22:1) for high erucic acid rapeseed.

Figure 4: Comparison between gas chromatography data and NIR predicted values of erucic acid (C22:1) for high erucic acid condiment mustard.

The graphs shown in Figures 1 – 4 display the range of values and the agreement between the calibration and the predicted data. There is a tendency in these graphs to exhibit some clustering, however, it should be noted that the selection of these calibration samples was based upon the uniqueness of their spectra and not on the spread of constituent values.

Results of this study were similar to those obtained by Panford and deMan (1990) and Reinhardt and Röbbelen (1991), however, their calibrations were based on wavelength calculations. The use of full spectrum information in principle component space as used by Golebiowski et al (1995) allows for a relatively easy test for spectral similarity between the calibration set of spectra and new incoming spectra.

CONCLUSION:

Brassica cultivars of different botanical origin and quality exhibit heterogeneous NIR spectra. Calibrations that allow for the simultaneous determination of the eight major fatty acids in these oilseeds were developed for each separate group. While FAME-GC remains the primary method for developing these calibrations, NIR analysis can be used to screen thousands of samples, saving a considerable amount of time and effort in the laboratory.

REFERENCE:

1. Daun, J.K and Mazur, P.B. ‘Preparation of Methyl Esters of Vegetable oil using Sodium Methoxide derivitisation: A brief description’. Journal of American Oil Chemists Society, Vol. 60, no. 10, pp.1751-1753, 1983.

2. Panford, J.A. and deMan, J.M, ‘Determination of Oil Content by NIR: Influence of Fatty Acid Composition on Wavelength Selection,’ Journal of American Oil Chemists Society, Vol. 67, no. 8, pp 473-482, 1990.

3. Reinhardt, T.C. and Röbbelen, G. ‘Quantitative Analysis of Fatty Acids in Intact Rapeseed by Near Infrared Reflectance Spectroscopy,’ GCIRC 1991 Congress, pp 1380-1384.

4. Sato, T., S. Kawano and M. Iwamoto, ‘Near Infrared Spectral Patterns of Fatty Analysis from Fats and Oils,’ Journal of American Oil Chemists Society, Vol. 68, no. 11, pp 827-833, 1991.

5. Golebiowski, T, Filip, M.L, Harman, C.F. and Pallot, T. ‘Routine Analysis of Fatty Acids in Whole Seed of Canola by NIR,’ 10^th Australian Research Assembly on Brassicas, September 1995, pp 87-89.

6. Shenk, J.S. and Westerhaus, M.O. Routine Operation, Calibration Development and Network System Management Manual for Near Infrared Instruments, Ver. 3.1, Infrasoft International, 1995.

7. Velasco, L, Fernández-Martínez, J.M. and De Haro, A. ‘The Applicability of NIRS for Estimating Multiple Seed Quality Components in Ethiopian Mustard,’ Rapeseed: today and tomorrow, Vol. 3, 9^th International Rapeseed Congress, July 1995, pp 867-869.