Published online 8 September 2006
Published in Crop Sci 46:2121-2126 (2006)
© 2006 Crop Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
CROP BREEDING & GENETICS
Predicting Oleic and Linoleic Acid Content of Single Peanut Seeds using Near-Infrared Reflectance Spectroscopy
Barry L. Tillmana,*,
Daniel W. Gorbeta and
George Personb
a Univ. of Florida, North Florida Research and Education Center, 3925 Hwy. 71, Marianna, FL 32446
b Agronomy Dep., Univ. of Florida, 2062 McCarty Hall, Gainesville, FL 32611
* Corresponding author (btillman{at}ufl.edu)
 |
ABSTRACT
|
|---|
One objective of peanut breeders is to develop cultivars with elevated oleic acid content (>740 g kg–1). Testing single seeds for oleic acid content is possible using gas chromatography (GC), but it is time-consuming and requires cutting a portion of the seed which could reduce germination. Using single, intact peanut seeds, we developed a near-infrared reflectance spectroscopy (NIR) calibration equation relating oleic acid measured with GC to oleic acid predicted by NIR. The slope of the regression line of oleic acid measured with GC on oleic acid predicted by NIR was 1.01 g kg–1 (P > t < .0001) and the intercept was not different from zero. An independent set of 95 peanut seeds was used to validate the NIR calibration. The slope of the regression line was 1.01 g kg–1 (P > t < .0001) and the intercept was not different from zero. By selecting seeds with at least 700 g kg–1 NIR predicted oleic acid content, only four of the 43 seeds (validation set) with elevated oleic acid content were misclassified by NIR, and none with normal oleic acid content were misidentified. Results were similar for linoleic acid content. This research shows that NIR prediction of oleic acid and linoleic acid using intact peanut seeds is accurate and rapid and should be especially useful for early generation screening.
Abbreviations: NIR, near-infrared reflectance spectroscopy • GC, gas chromatography • PLS, principal least squares • RMSECV, root mean squared error of cross-validation • RMSEP, root mean squared error of prediction
|
INTRODUCTION
|
|---|
OLEIC ACID (a monounsaturated fatty acid) content of oilseeds has important implications for product performance and consumer health. Typically, peanut seeds contain about 500 to 600 g kg–1 oleic acid. Peanut genotypes containing oleic acid content of greater than 780 g kg–1 have been discovered and cultivars with this trait have been developed and released for commercialization (Norden et al., 1987; Gorbet and Knauft, 1997, 2000; Branch, 2000). In clinical trials, regular peanut consumption has been shown to improve dietary and biochemical elements that are associated with reduced risk of cardiovascular disease (Alper and Mattes, 2003; Kris-Etherton et al., 1999). The peanut seeds used in these clinical trials were not likely to have elevated oleic acid content, although their oleic acid content is thought to be a major contributor to the observed differences. Another study at the University of Florida used peanut seeds with elevated oleic acid and found a reduction in the blood cholesterol in hyperchloresterolemic women (O'Byrne et al., 1993). When roasted in the shell, peanut seeds with elevated oleic acid content (>740 g kg–1) and reduced linoleic acid content (<80 g kg–1) have up to 8 times greater shelf-life than peanut seeds with normal levels of oleic acid and linoleic acid content (500–650 g kg–1 oleic acid and 200–300 g kg–1 linoleic acid content) (Mozingo et al., 2004). A similar study of the shelf-life of peanut oil showed the oil with elevated oleic acid had a 14-fold advantage over peanut oil with normal oleic acid (O'Keefe et al., 1993). When swine were fed peanut seed high in oleic acid content, monounsaturated fatty acids in the swine fats increased and polyunsaturated fatty acids decreased, compared to a control diet of normal oleic acid peanut (Myer et al., 1992).
Given the potential importance of elevated oleic acid content in peanuts, breeding programs measure oleic acid content in thousands of samples each year. Historically, fatty acid content of seed is measured using GC and most methods require destruction of the sample. However, researchers have utilized NIR to replace GC, for measurement of oleic acid and linoleic acid. Velasco et al. (1999) developed NIR calibration equations for single achenes of sunflower that were highly predictive of oleic acid and linoleic acid measured by GC. Similarly NIR has been used to predict oleic acid content of ground and whole soybean [Glycine max (L.) Merr.] seed (Pazdernik et al., 1997) and intact rapeseed (Brassica napus L.) (Velasco and Becker, 1998). The University of Florida peanut breeding program uses GC to measure oleic acid and linoleic acid content in seeds of breeding lines. Each sample requires about 5 min to prepare and 20 min for GC analysis. Sample preparation requires chopping the seeds so that testing of segregating material is impossible. Other GC methods utilize part of the peanut seed so that planting seeds is possible. With this method, injury to the seeds causes only minor reduction in germination, but the method is still time-consuming compared to NIR (Zeile et al., 1993). The purpose of this research was to develop predictive equations for NIR that would allow rapid, nondestructive testing of oleic acid and linoleic acid content of peanut seeds.
|
MATERIALS AND METHODS
|
|---|
Seeds of 89 peanut cultivars and breeding lines (132 total seeds) were scanned by NIR and subsequently tested for oleic acid and linoleic acid using GC. The GC method we used was described by Metcalf et al. (1966). Spectral data for each peanut seed were obtained from a ThermoNicolet Industrial Solutions (Fitchburg, WI) Nexus 670 FT-IR scanning monochronometer equipped with a NearIR UpDrift Smart Accessory. Each individual seed was scanned 8 times with a mirror velocity of 1.2659 and an aperture of 10. Wavelengths were 1000 to 2500 nm. Using GC, measurements of oleic acid and linoleic acid were matched with their respective NIR scan, and a calibration equation was developed with ThermoNicolet software TQ Analyst version 6.1.1.356. The software analysis option used to develop the equations was principal least squares (PLS) with a multiplicative signal correction pathlength. The second derivative of the spectra is processed with the Norris derivative filter set to a segment length of 9 and a gap between segments of 9. Data normalization used the variance scaling technique. A simple fit value algorithm (measured from zero) was used and there were 8 PLS factors. The software conducted two model validation analyses. First, a cross-validation analysis was conducted in which each sample was left out and error was evaluated statistically on prediction of the left out samples, resulting in the root mean squared error of cross-validation (RMSECV). Second, a certain number of samples (20 in this case) are designated as validation samples and the root mean squared error of prediction (RMSEP) was calculated. The calibration dataset was subjected to regression analysis using SAS to statistically describe the relationship between the values predicted by NIR and those obtained by GC (SAS Institute, 2000). A separate set of 95 individual seeds from 19 different genotypes was subjected to fatty acid testing using NIR and GC for model/method validation. The NIR-predicted and GC-measured oleic acid and linoleic acid values from the validation dataset were then subjected to regression analysis using PROC REG of SAS.
|
RESULTS AND DISCUSSION
|
|---|
Calibration and Cross Validation
The NIR calibration equations were highly predictive of oleic acid and linoleic acid measured with GC (Table 1). The standard error of the mean (RMSEC) was 15 g kg–1 for oleic acid and 13 g kg–1 for linoleic acid, or 2 and 11% of the mean, respectively (Table 1). The root mean squared error of cross-validation (RMSECV) was 33 g kg–1 for oleic acid and 27 g kg–1 for linoleic acid or about 5 and 22% of the mean, respectively. The RMSEP was 20 g kg–1 for oleic acid and 19 g kg–1 for linoleic acid, or 3 and 16% of the mean, respectively.
View this table:
[in this window]
[in a new window]
|
Table 1. Statistics from calibration equations developed using the software program ThermoNicolet TQAnalyst version 6.1.1.356 to predict oleic and linoleic acid of intact peanut seed with near-infrared reflectance spectroscopy (NIR).
|
|
The slope of the simple linear regression of GC observations on NIR predictions was 1.01 g kg–1 (p > |t| < 0.0001) for both oleic acid and linoleic acid and the intercept was not different than zero (Fig. 1
). Slight trends were noted in a plot of the residuals against the NIR-predicted values for both oleic acid and linoleic acid. In both cases the trend occurred at the extremes (greater than 740 g kg–1 oleic acid and less than 100 g kg–1 linoleic acid) which include primarily seeds with elevated oleic acid (Fig. 2
). This is due to the fact that the range of oleic acid and linoleic acid measured by GC at the extremes was small (about 50 g kg–1) in comparison to the range of oleic acid and linoleic acid predicted by NIR (about 125 g kg–1). The same phenomenon appears to have occurred when NIR was used to predict oleic acid and linoleic acid of intact sunflower achenes (Velasco et al., 1999), although residual plots were not presented. Residuals appeared to be normally distributed within the normal range of oleic acid and linoleic acid content of 400 to 700 g kg–1 oleic acid and 80 to 400 g kg–1 linoleic acid (Fig. 2).
View larger version (43K):
[in this window]
[in a new window]
|
Fig. 1. Calibration equation for prediction of oleic acid (A) and linoleic acid (B) content in single intact peanut seeds from near-infrared reflectance (NIR) spectroscopy.
|
|
View larger version (104K):
[in this window]
[in a new window]
|
Fig. 2. Residuals from regression of NIR-predicted oleic acid (A) and linoleic acid (B) from single intact peanut seeds plotted against the NIR-predicted value.
|
|
External Validation
The slope of the regression line of oleic acid and linoleic acid measured by GC on that predicted by NIR was 1.01 g kg–1 (P > |t| < 0.0001) and 1.00 g kg–1, respectively (Fig. 3
). The intercept was not different from zero for either oleic acid or linoleic acid. The R2 values were lower than those from calibration (0.98 for both fatty acids) at 0.84 for oleic acid and 0.85 for linoleic acid.
View larger version (46K):
[in this window]
[in a new window]
|
Fig. 3. External validation of a calibration equation for prediction of oleic acid (A) and linoleic acid (B) content in single intact peanut seeds using near-infrared reflectance (NIR) spectroscopy.
|
|
The NIR calibration equation accurately identified over 90% of the seeds with elevated oleic acid and reduced linoleic acid in the external validation dataset (Table 2). All of the seeds with elevated oleic acid or reduced linoleic acid that were misclassified by NIR were identified as having normal oleic acid and linoleic acid in the range of 650 to 700 g kg–1. It is preferential that none of the misclassified seeds were falsely identified as having elevated oleic acid or reduced linoleic acid. This is because there are a relatively small number of samples in the oleic acid range of 650 to 700 g kg–1 that would require retesting on GC, compared to eliminating only the normal oleic acid seeds with NIR and retesting all seeds predicted to have elevated oleic acid.
View this table:
[in this window]
[in a new window]
|
Table 2. Statistics from prediction of oleic and linoleic acid of single peanut seeds using a near-infrared reflectance (NIR) spectroscopy calibration and performance of the method based on 43 seeds verified to have elevated oleic acid and reduced linoleic acid.
|
|
|
CONCLUSIONS
|
|---|
Our results show that NIR can be utilized to predict the oleic and linoleic acid concentration of single peanut seeds. From a set of 95 individual seeds not used in the calibration, the method accurately identified 90% of the seeds with elevated oleic acid content. Of the 10% that were misclassified, all were predicted to have normal oleic acid content, and no seeds with normal oleic acid content were misclassified as having elevated oleic acid content. The NIR method is also rapid. Prediction of the fatty acid composition of a single seed with NIR requires no more than 30 sec compared to 20 to 30 min using GC. Using NIR to predict oleic and linoleic acid content of individual peanut seeds should be valuable for peanut breeding programs.
Received for publication January 18, 2006.
|
REFERENCES
|
|---|
- Alper, C.M., and R.D. Mattes. 2003. Peanut consumption improves indices of cardiovascular disease risk in healthy adults. J. Am. Coll. Nutr. 22:133–141.[Abstract/Free Full Text]
- Branch, W.D. 2000. Registration of ‘Georgia Hi-O/L’ peanut. Crop Sci. 40:1823–1824.
- Gorbet, D.W., and D.A. Knauft. 1997. Registration of ‘SunOleic 95R’ peanut. Crop Sci. 37:1392.[Free Full Text]
- Gorbet, D.W., and D.A. Knauft. 2000. Registration of ‘SunOleic 97R’ peanut. Crop Sci. 40:1190–1191.[ISI]
- Kris-Etherton, P.M., T.A. Pearson, Y. Wan, R.L. Hargrove, K. Moriarty, V. Fishell, and T.D. Etherton. 1999. High-monounsaturated fatty acid diets lower both plasma cholesterol and triacylglycerol concentrations. Am. J. Clin. Nutr. 10:1009–1015.
- Metcalf, L.D., A.A. Schmitz, and J.K. Pelka. 1966. Rapid preparation of fatty acid esters from lipids for gas chromatographic analysis. Anal. Chem. 38:514–515.[CrossRef]
- Mozingo, R.W., S.F. O'Keefe, T.H. Sanders, and K.W. Hendrix. 2004. Improving shelf life of roasted and salted inshell peanuts using high oleic fatty acid chemistry. Peanut Sci. 31:40–45.
- Myer, R.O., D.D. Johnson, D.A. Knauft, D.W. Gorbet, J.H. Brendemuhl, and W.R. Walker. 1992. Effect of feeding high-oleic-acid peanuts to growing-finishing swine on resulting carcass fatty acid profile and on carcass and meat quality characteristics. J. Anim. Sci. 70:3734–3741.[Abstract]
- Norden, A.J., D.W. Gorbet, D.A. Knauft, and C.T. Young. 1987. Variability in oil quality among peanut genotypes in the Florida breeding program. Peanut Sci. 14:7–11.
- O'Byrne, D.J., R.B. Shireman, and D.A. Knauft. 1993. The effects of a low fat/high oleic acid diet on lipoprotein in postmenopausal hypercholesterolemic women. Proc. 84th Amer. Oil Chemists Soc. Ann. Meeting (AOCS).
- O'Keefe, S.F., V.A. Wiley, and D.A. Knauft. 1993. Comparison of oxidation stability of high- and normal-oleic peanut oils. J. Am. Oil Chem. Soc. 70:489–492.[CrossRef][ISI]
- Pazdernik, D.L., A.S. Killam, and J.H. Orf. 1997. Analysis of amino and fatty acid composition in soybean seed, using near-infrared reflectance spectroscopy. Agron. J. 89:679–685.[Abstract/Free Full Text]
- SAS Institute. 2000. Statistical Analysis Software for Windows. Version 8.1. SAS Institute, Cary, NC.
- Velasco, L., B. Perez-Vich, and J.M. Fernandez-Martinez. 1999. Nondestructive screening for oleic and linoleic acid in single sunflower achenes by near-infrared reflectance spectroscopy. Crop Sci. 39:219–222.[Abstract/Free Full Text]
- Velasco, L., and H.C. Becker. 1998. Estimating the fatty acid composition of the oil in intact-seed rapeseed (Brassica napus L.) by near-infrared reflectance spectroscopy. Euphytica 101:221–230.[CrossRef]
- Zeile, W.L., D.A. Knauft, and C.B. Kelly. 1993. A rapid non-destructive technique for fatty acid determination in individual peanut seed. Peanut Sci. 20:9–11.
TOP
ABSTRACT
INTRODUCTION
