Abstract
Food authentication is one of the major problems faced by beef industries in both Australia and around the world in particular authenticating production systems for beef products as consumers are increasingly demanding more information on how their meat is produced. A portable rapid objective measure of predicting production system was developed as a result of this expectation of consumers of true to label products. Raman spectroscopy was chosen for investigation as it is able to provide a chemical fingerprint of the sample that should be able to detect changes in feed, provides a rapid result, does not require specialised training for operation and is commercially available in a portable handheld device.
The current method of authenticating the production system of beef is dependent on audits and reliant on producers following the requirements set by processors, which vary for individual grain and grass-fed brands. Maintaining transparency and trust in the supply chain for grass fed beef represents a significant cost to the industry in the form of auditing and an even greater potential cost if there is a failure in the auditing process. The overall aim of this
research will be to test the viability of Raman Spectroscopy to accurately differentiate between production systems. This thesis presents an evaluation of Raman Spectroscopy in combination with a variety of chemometric modelling techniques, to accurately classify the production system of beef carcases from across Australia.
Raman spectra was collected at 24 h post mortem from the subcutaneous fat over the brisket of beef carcases, using a 785 nm handheld portable Raman device. The beef cattle were sourced from a variety of production systems from across Australia including 100 day grain fed, verified grass fed, northern long term grain fed, northern short term grain fed, northern grass fed, northern grass fed with supplementation, southern long term grain fed, southern short term grain fed, southern grass fed, and southern grass fed with supplementation. The ability for Raman spectra to discriminate between the production systems was assessed utilising Principal Components Analysis (PCA) and Partial Least Square Discriminant Analysis (PLS-DA). These chemometric techniques were also utilised to asses the fatty acid composition of the subcutaneous fat assessed by fatty acid methylation extraction using gas chromatography with a flame ionisation detector (GC-FID).
Overall the classification of beef production systems based on Raman spectroscopy measures of fat composition showed promising results with the southern production systems providing the best accuracy of 95- 98 % for discriminating between different levels of grain and grass feeding. The spectral features of note included peaks at 1301 cm−1, 1206 cm−1, 1440 cm−1, 1506 cm−1, 1626 cm−1, 1658 cm−1, 1686 cm−1 and 1803 cm−1.
The northern data performed best when the model was simplified, developed utilising a two-class grain vs grass system, this resulted in an increase in accuracy from 59-92 % to an overall model accuracy of 94 %. The spectral features utilised for the discrimination of the northern data included 880-920 cm -1, 1080-1100 cm-1, 1206 cm-1, 1301 cm-1, 1440 cm-1, 1650- 1660 cm-1 and 1745-1760 cm-1. Assessment of these wavelengths indicates the differences are due to the chemical structures C-C, CH, CH2, CH3, C-O, C=O and C=C. These structures and location in the spectra have been linked consistently to the fatty acids and overall lipids. Due to this the fatty acid composition of the subcutaneous fat was assessed. When compared with
the Raman spectra, models using the fatty acid data developed using PLS-DA was comparable or outperformed the Raman spectra when assessing the individual production systems, northern long term grain fed, northern short term grain fed, northern grass fed, northern grass fed with supplementation, southern long term grain fed, southern short term grain fed, southern grass fed, and southern grass fed with supplementation, in an eight class model. When predicting grain vs grass feeding the results obtained from the fatty acid data and Raman spectra were comparable. Whilst the fatty acid data provided a more accurate result, this method of measurement is not viable for industry, due to the length of time required in obtaining a result, the use of caustic reagents and the high cost. As such Raman spectroscopy, with comparable results to the fatty acid data, is a piece of technology that shows promise for industry application as a classification model to predict grain and grass-fed carcases. Hence, Raman spectroscopy is an ideal objective method to classify beef carcases according to production system.
The current method of authenticating the production system of beef is dependent on audits and reliant on producers following the requirements set by processors, which vary for individual grain and grass-fed brands. Maintaining transparency and trust in the supply chain for grass fed beef represents a significant cost to the industry in the form of auditing and an even greater potential cost if there is a failure in the auditing process. The overall aim of this
research will be to test the viability of Raman Spectroscopy to accurately differentiate between production systems. This thesis presents an evaluation of Raman Spectroscopy in combination with a variety of chemometric modelling techniques, to accurately classify the production system of beef carcases from across Australia.
Raman spectra was collected at 24 h post mortem from the subcutaneous fat over the brisket of beef carcases, using a 785 nm handheld portable Raman device. The beef cattle were sourced from a variety of production systems from across Australia including 100 day grain fed, verified grass fed, northern long term grain fed, northern short term grain fed, northern grass fed, northern grass fed with supplementation, southern long term grain fed, southern short term grain fed, southern grass fed, and southern grass fed with supplementation. The ability for Raman spectra to discriminate between the production systems was assessed utilising Principal Components Analysis (PCA) and Partial Least Square Discriminant Analysis (PLS-DA). These chemometric techniques were also utilised to asses the fatty acid composition of the subcutaneous fat assessed by fatty acid methylation extraction using gas chromatography with a flame ionisation detector (GC-FID).
Overall the classification of beef production systems based on Raman spectroscopy measures of fat composition showed promising results with the southern production systems providing the best accuracy of 95- 98 % for discriminating between different levels of grain and grass feeding. The spectral features of note included peaks at 1301 cm−1, 1206 cm−1, 1440 cm−1, 1506 cm−1, 1626 cm−1, 1658 cm−1, 1686 cm−1 and 1803 cm−1.
The northern data performed best when the model was simplified, developed utilising a two-class grain vs grass system, this resulted in an increase in accuracy from 59-92 % to an overall model accuracy of 94 %. The spectral features utilised for the discrimination of the northern data included 880-920 cm -1, 1080-1100 cm-1, 1206 cm-1, 1301 cm-1, 1440 cm-1, 1650- 1660 cm-1 and 1745-1760 cm-1. Assessment of these wavelengths indicates the differences are due to the chemical structures C-C, CH, CH2, CH3, C-O, C=O and C=C. These structures and location in the spectra have been linked consistently to the fatty acids and overall lipids. Due to this the fatty acid composition of the subcutaneous fat was assessed. When compared with
the Raman spectra, models using the fatty acid data developed using PLS-DA was comparable or outperformed the Raman spectra when assessing the individual production systems, northern long term grain fed, northern short term grain fed, northern grass fed, northern grass fed with supplementation, southern long term grain fed, southern short term grain fed, southern grass fed, and southern grass fed with supplementation, in an eight class model. When predicting grain vs grass feeding the results obtained from the fatty acid data and Raman spectra were comparable. Whilst the fatty acid data provided a more accurate result, this method of measurement is not viable for industry, due to the length of time required in obtaining a result, the use of caustic reagents and the high cost. As such Raman spectroscopy, with comparable results to the fatty acid data, is a piece of technology that shows promise for industry application as a classification model to predict grain and grass-fed carcases. Hence, Raman spectroscopy is an ideal objective method to classify beef carcases according to production system.
| Original language | English |
|---|---|
| Qualification | Doctor of Philosophy |
| Awarding Institution |
|
| Supervisors/Advisors |
|
| Place of Publication | Australia |
| Publisher | |
| Publication status | Published - 2021 |
Grant Number
- Raman Spectroscopy
- Beef Cattle
- Fatty Acid Composition
- Grass fed beef
- Grain fed beef