Determination and Prediction of Fumonisin Contamination in Maize by Surface–Enhanced Raman Spectroscopy (SERS)

Abstract

The potential and limitation of surface-enhanced Raman spectroscopy (SERS) method was investigated to develop an accelerated spectroscopic method as an alternative analytical technique to commonly used wet chemical methods for fumonisin analysis in maize. SERS spectral difference among groups of ground maize samples with different concentrations of fumonisins more clearly reflected the level of fumonisin contamination and its effect on physicochemical properties of ground maize samples than conventional Raman spectral difference. In general, chemometric classification models exhibited moderately acceptable correct classification rates (68.0–100.0 % for training dataset and 58.8–85.3 % for validation dataset) and no or little false–negative error. The k-nearest neighbor models applied to validation dataset slightly outperformed over other classification models, showing correct classification rates of 70.6–79.4 %. Chemometric quantification models using validation dataset also yielded a good predictive power and ability, showing satisfactory regression quality (slope = 0.902–1.096), high coefficient of determination (r 2 = 0.825–0.940), and low root-mean-square error of prediction (RMSEP = 11.162–19.954 mg/kg), with no statistical significant difference with the reference value. The multiple linear regression models showed better quality of linear regression (slope = 0.902–1.076), stronger correlation coefficient (r = 0.948–0.969), and higher predictive accuracy (r 2 = 0.900–0.940) than other quantification models. The proposed SERS method would be a suitable and convenient analytical tool with a great potential for improvement in qualitative and quantitative characterization of fumonisins in maize, serving as a valuable screening tool for maize samples contaminated with fumonisins at a point of sampling.