Loading...
Proximate composition and fatty acid profile of South African sardines (Sardinops sagax ocellatus) using conventional techniques and near infrared (NIR) spectroscopy
Author(s)
Phike, Ziyanda
Date Issued
2020
Type
Thesis
Publisher
Cape Peninsula University of Technology
Abstract
Fish and fishery products form an essential part of a healthy human diet due to the high-quality proteins and omega-3 fatty acids associated with it. The South African (SA) sardine (Sardinops sagax ocellatus) is commercially important in South Africa’s purse seine fishery and the fish canning industry. Reasons being sardines are an important source of high-quality proteins and fatty acids and therefore used as a food source for humans and pets. Moreover, sardine fishery offers employment to a large SA workforce. The growing consumption of sardines has led to the fish canning industry investigating effective instrumental techniques of evaluating the quality of the respective products.
Multivariate data analysis, in particular principal component analysis (PCA), was applied on acquired near infrared (NIR) spectra to investigate the correlation between SA sardines based on morphophysiological properties. Sample sets of whole (n=116), and homogenised (n=97) sardine samples were investigated based on sex, fat tage and gonad stage, using PCA. For the whole fish sample set, groupings of sex, fat and gonad stages were distinguishable. Homogenised fish samples could not be classified based on sex and gonad stages. Only the fat stages could be differentiated when using the homogenised fish samples.
Homogenized sardine samples were scanned by NIR. Sub samples were used to determine percentage moisture, ash, protein and fat using conventional methods. The data sets from the NIR scans and from the chemical analysis were combined to develop calibration models to allow prediction of proximate composition. These results were used as reference values to build partial least square (PLS) regression models using the NIR data-spectra. Prediction statistics for the respective validation sets were: R2v = 0.22, SEP = 3.14%, RPD 1.15 for moisture, R2v = 0.24, SEP = 0.056%, RPD 1.07 for ash, R2v = 0.61, SEP = 2.46%, RPD 1.47 for fat and R2v = 0.22, SEP = 3.41%, RPD 1.07 for protein. The results indicated that no reasonable predictions could be made at the fast NIR scan rate.
Additionally, the fatty acid composition of sardines was obtained using a conventional gas chromatography (GC) method. These results were used to build PLS regression models in combination with NIR spectra acquired from the respective sample sets. The developed PLS models resulted in poor predictions for saturated fatty acids (SFA): R2v = 0.48, SEP = 0.45% and RPD = 0.65 and polyunsaturated fatty acids (PUFA): R2v = 0.48, SEP = 0.44% and RPD = 1.47. Monounsaturated fatty acids (MUFA) had very poor prediction results with very little correlation found between the predicted and reference values (R2v = 0.38, SEC = 2.33). Likewise, total fatty acids resulted in poor predictions (R2v = 0.37, SEC = 2.03). NIR spectroscopy has potential to be used as an analytical technique to predict proximate and fatty acid composition of fish, however further improvement is still necessary before it can be considered for industrial use.
Multivariate data analysis, in particular principal component analysis (PCA), was applied on acquired near infrared (NIR) spectra to investigate the correlation between SA sardines based on morphophysiological properties. Sample sets of whole (n=116), and homogenised (n=97) sardine samples were investigated based on sex, fat tage and gonad stage, using PCA. For the whole fish sample set, groupings of sex, fat and gonad stages were distinguishable. Homogenised fish samples could not be classified based on sex and gonad stages. Only the fat stages could be differentiated when using the homogenised fish samples.
Homogenized sardine samples were scanned by NIR. Sub samples were used to determine percentage moisture, ash, protein and fat using conventional methods. The data sets from the NIR scans and from the chemical analysis were combined to develop calibration models to allow prediction of proximate composition. These results were used as reference values to build partial least square (PLS) regression models using the NIR data-spectra. Prediction statistics for the respective validation sets were: R2v = 0.22, SEP = 3.14%, RPD 1.15 for moisture, R2v = 0.24, SEP = 0.056%, RPD 1.07 for ash, R2v = 0.61, SEP = 2.46%, RPD 1.47 for fat and R2v = 0.22, SEP = 3.41%, RPD 1.07 for protein. The results indicated that no reasonable predictions could be made at the fast NIR scan rate.
Additionally, the fatty acid composition of sardines was obtained using a conventional gas chromatography (GC) method. These results were used to build PLS regression models in combination with NIR spectra acquired from the respective sample sets. The developed PLS models resulted in poor predictions for saturated fatty acids (SFA): R2v = 0.48, SEP = 0.45% and RPD = 0.65 and polyunsaturated fatty acids (PUFA): R2v = 0.48, SEP = 0.44% and RPD = 1.47. Monounsaturated fatty acids (MUFA) had very poor prediction results with very little correlation found between the predicted and reference values (R2v = 0.38, SEC = 2.33). Likewise, total fatty acids resulted in poor predictions (R2v = 0.37, SEC = 2.03). NIR spectroscopy has potential to be used as an analytical technique to predict proximate and fatty acid composition of fish, however further improvement is still necessary before it can be considered for industrial use.
Additional information
Thesis (Master of Food Science and Technology)--Cape Peninsula University of Technology, 2020
File(s)![Thumbnail Image]()
Loading...
Name
Ziyanda_Phike_210211628.pdf
Size
2.97 MB
Format
Adobe PDF
Checksum
(MD5):0c457e56ba8ca1d2cfb643baf85d8c30