NIRS Calibration of Aflatoxin in Maize
Ross E. Darnell A F , Jagger J. Harvey B , Glen P. Fox C , Mary T. Fletcher C , James Wainaina D , Immaculate Wanjuki E and Warwick J. Turner EA DATA61, CSIRO, PO Box 2583, Brisbane, Qld 4001, Australia.
B Feed the Future Innovation Lab for the Reduction of Post-Harvest Loss, 103B Waters Hall, 1603 Old Claflin Place, Kansas State University, Manhattan, KS 66506, USA.
C Queensland Alliance for Agriculture and Food Innovation (QAAFI), The University of Queensland, Health and Food Sciences Precinct, Coopers Plains, Qld 4108, Australia.
D The University of Western Australia, Australian Research Council Centre of Excellence in Plant Energy Biology and School of Molecular Sciences, Crawley, Perth, WA 6009, Australia.
E BecA (Biosciences East Central Africa), PO Box 30709, Nairobi 00100, Kenya.
F Corresponding author. Email: ross.darnell@csiro.au
Australian Journal of Chemistry 71(11) 868-873 https://doi.org/10.1071/CH18316
Submitted: 2 July 2018 Accepted: 29 August 2018 Published: 13 September 2018
Abstract
The aim of this study is to determine the value of near-infrared spectroscopy (NIRS) as a diagnostic tool for aflatoxin contamination, specifically to rapidly predict levels of aflatoxin, either quantitatively or qualitatively, in ground maize. Maize was collected from inoculated field trials conducted across four sites in Kenya. Inoculated and uninoculated maize ears were harvested, milled, and prepared for NIRS scanning and wet chemistry-based aflatoxin quantification. Several statistical and machine learning techniques were compared. Absorbance at a single bandwidth explained 34 % of the variation in levels of aflatoxin using a regression model while a partial least-squares (PLS) method showed that NIR measurements could explain 42 % of the variation in aflatoxin levels. To compare various methods for their ability to classify samples with high (>100 ppb) levels of aflatoxin, various additional procedures were applied. The k-nearest neighbour classification method yielded sensitivity and specificity values of 0.75 and 0.52 respectively, compared with the support vector machine method with estimates of 0.81 and 0.68, whereas PLS could achieve values of 0.82 and 0.72 respectively. The corresponding false positive and false negative values are still unacceptable for NIRS to be used with confidence, as ~18 % of contaminated ground maize samples would be accepted and 28 % of good maize would be discarded or declared contaminated or downgraded. However, such calibrations could be useful in breeding programs without access to wet chemistry analysis, seeking to rank entries semiquantitatively.
References
[1] L. D. Williams, A. E. Glenn, C. W. Bacon, M. A. Smith, R. T. Riley, J. Agric. Food Chem. 2006, 54, 5694.| Crossref | GoogleScholarGoogle Scholar |
[2] J. J. Harvey, B. Gnonlonfin, M. Fletcher, G. Fox, S. Trowell, A. Berna, R. Nelson, R. Darnell, in Aflatoxins: Finding Solutions for Improved Food Safety (Eds L. J. Unnevehr, D. Grace) 2013, 2020 Vision Focus 20(19), pp. 1–2 (International Food Policy Research Institute: Washington, DC).
[3] B. G. Osborne, in Encyclopedia of Analytical Chemistry (Eds R. A. Meyers, R. J. McGorrin) 2006, pp. 1–14 (John Wiley & Sons, Ltd: Hoboken, NJ).
[4] G. Fox, M. Manley, J. Sci. Food Agric. 2014, 94, 174.
| Crossref | GoogleScholarGoogle Scholar |
[5] H. Yao, Z. Hruska, R. Kincaid, R. Brown, T. Cleveland, D. Bhatnagar, Food Addit. Contam. Part A 2010, 27, 701.
| Crossref | GoogleScholarGoogle Scholar |
[6] T. C. Pearson, D. Brabec, S. Haley, Sens. Instrum. Food Qual. Saf. 2008, 2, 280.
| Crossref | GoogleScholarGoogle Scholar |
[7] S. Wold, M. Sjöström, L. Eriksson, Chemom. Intell. Lab. Syst. 2001, 58, 109.
| Crossref | GoogleScholarGoogle Scholar |
[8] B.-H. Mevik, R. Wehrens, J. Stat. Softw. 2007, 18, 1.
[9] W. J. Turner, M. Fletcher, Y. L. Pang, L. Dora, G. P. Fox, R. Darnell, J. J. W. Harvey, in 46th Annual AIFST Convention 2013, Australian Institute of Food Science and Technology.
[10] M. E. Benvenuti, A. J. Di Gioia, Rapid Analysis of Aflatoxins Without Derivatization Using Ultra-Performance Liquid Chromatography and Fluorescence Detection. Waters Application Note. Available at: http://www.waters.com/webassets/cms/library/docs/720002644en.pdf (accessed 7 August 2018).
[11] R Core Team, R: A Language and Environment for Statistical Computing 2014 (R Foundation for Statistical Computing: Vienna).
[12] P. Brown, J. R. Stat. Soc. B 1982, 44, 287.
[13] S. Wold, H. Martens, H. Wold, in Matrix Pencils (Eds A. Ruhe, B. Kagstrom) 1983, Lecture Notes in Mathematics Vol. 973, pp. 286–293 (Springer Verlag: Heidelberg).
[14] B.-H. Mevik, R. Wehrens, K. H. Liland, Partial Least Squares and Principal Component Regression 2015. Available at https://CRAN.R-project.org/package=pls
[15] C.-W. Hsu, C.-C. Chang, C.-J. Lin, A Practical Guide to Support Vector Classification 2016 (Department of Computer Science, National Taiwan University: Taipei).
[16] B. D. Ripley, Pattern Recognition and Neural Networks 2007 (Cambridge University Press: Cambridge).
