抄録
A rapid method for predicting total aflatoxin (aflatoxins B1, B2, G1, and G2) in nutmeg extract was developed using fluorescence fingerprint (FF) and partial least squares (PLS) regression. FF is also known as excitation-emission matrix, which is a series of fluorescence spectra acquired by scanning an excitation wavelength. Nutmeg extract was artificially spiked with an aflatoxin reagent. The FF of spiked nutmeg extract was measured with a fluorescence spectrometer. The FF data was preprocessed to remove signals not related to fluorescence. After preprocessing, 1428 out of 5041 fluorescence intensities remained. They were set as explanatory variables for PLS regression. Then, total aflatoxin concentration was set as the response variable. Eleven out of 21 samples were used as the calibration dataset; the remaining ten were used as the validation dataset. Three latent variables were used to develop the ideal PLS model by cross validation. R2 was 0.993 and SEC was 0.2 μg/L for the calibration dataset. Significant correlations were observed between the actual and predicted values for the validation dataset, with R2 of 0.773 and SEP of 1.0 μg/L. The PLS regression coefficient, which shows the degree of contribution of each wavelength to the model, indicated that the prediction was mainly based on the fluorescence of aflatoxin itself.
