Multivariate method for prediction of fumonisins B1 and B2 and zearalenone in Brazilian maize using Near Infrared Spectroscopy (NIR)

• Published in: PloS one 2021-01, Vol.16 (1), p.e0244957
• Main Authors: Tyska, Denize, Mallmann, Adriano Olnei, Vidal, Juliano Kobs, Almeida, Carlos Alberto Araújo de, Gressler, Luciane Tourem, Mallmann, Carlos Augusto
• Format: Article
• Language: English
• Subjects: Algorithms; Animals; Barley; Biology and Life Sciences; Brazil; Cereals; Chromatography; Contamination; Corn; Correlation coefficient; Correlation coefficients; Food contamination & poisoning; Food Contamination - analysis; Fumonisin B1; Fumonisin B2; Fumonisins - analysis; Fungi; Infrared spectroscopy; Laboratories; Liquid chromatography; Mass spectrometry; Mass spectroscopy; Medicine and Health Sciences; Methods; Mycotoxins; Mycotoxins - analysis; Near infrared radiation; Near infrared spectroscopy; Nondestructive testing; People and places; Physical Sciences; Predictions; Raw materials; Regression analysis; Regression models; Research and Analysis Methods; Spectroscopy, Near-Infrared; Spectrum analysis; Standard error; Toxins; Zea mays - chemistry; Zearalenone; Zearalenone - analysis
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0244957

Fumonisins (FBs) and zearalenone (ZEN) are mycotoxins which occur naturally in grains and cereals, especially maize, causing negative effects on animals and humans. Along with the need for constant monitoring, there is a growing demand for rapid, non-destructive methods. Among these, Near Infrared Spectroscopy (NIR) has made great headway for being an easy-to-use technology. NIR was applied in the present research to quantify the contamination level of total FBs, i.e., fumonisin B1+fumonisin B2 (FB1+FB2), and ZEN in Brazilian maize. From a total of six hundred and seventy-six samples, 236 were analyzed for FBs and 440 for ZEN. Three regression models were defined: one with 18 principal components (PCs) for FB1, one with 10 PCs for FB2, and one with 7 PCs for ZEN. Partial least square regression algorithm with full cross-validation was applied as internal validation. External validation was performed with 200 unknown samples (100 for FBs and 100 for ZEN). Correlation coefficient (R), determination coefficient (R2), root mean square error of prediction (RMSEP), standard error of prediction (SEP) and residual prediction deviation (RPD) for FBs and ZEN were, respectively: 0.809 and 0.991; 0.899 and 0.984; 659 and 69.4; 682 and 69.8; and 3.33 and 2.71. No significant difference was observed between predicted values using NIR and reference values obtained by Liquid Chromatography Coupled to Tandem Mass Spectrometry (LC-MS/MS), thus indicating the suitability of NIR to rapidly analyze a large numbers of maize samples for FBs and ZEN contamination. The external validation confirmed a fair potential of the model in predicting FB1+FB2 and ZEN concentration. This is the first study providing scientific knowledge on the determination of FBs and ZEN in Brazilian maize samples using NIR, which is confirmed as a reliable alternative methodology for the analysis of such toxins.