| Texto completo | |
| Autor(es): |
Carames, E. T. S.
;
Carames, J. G. S.
;
Moraes-Neto, V. F.
;
Bertozzi, B. G.
;
Valderrama, P.
;
Correa, B.
;
Rocha, L. O.
;
Pallone, J. A. L.
Número total de Autores: 8
|
| Tipo de documento: | Artigo Científico |
| Fonte: | Microchemical Journal; v. 219, p. 8-pg., 2025-11-23. |
| Resumo | |
The negative impact of the Fusarium sambucinum species complex (FSAMSC) on grass cultivars such as barley is undeniable, due to its capacity to cause plant diseases and produce mycotoxins, mostly trichothecenes and zearalenone. To address this, the development of rapid analytical methods to inform management and control strategies is essential. This study evaluates ultraviolet-visible (UV-Vis) spectroscopy coupled with chemometric tools for classifying four FSAMSC species (F. poae, F. graminearum, F. cortaderiae, and F. meridionale) represented by 46 strains isolated from Brazilian barley. It was compared the performance of four models: Partial Least Squares Discriminant Analysis (PLS-DA), Support Vector for Classification (SVC), Random Forest (RF), and Extreme Gradient Boosting (XGBoost), based on prediction accuracy and computational efficiency. The XGBoost model achieved the highest mean accuracy (0.92), closely followed by SVC (0.90). However, SVC required only 5.92 min of processing, which is markedly shorter than the 1726 min demanded by XGBoost modeling. This substantial difference highlights XGBoost's higher computational cost and, consequently, greater energy consumption, with 0.000633 kWh for SVC versus 0.217528 kWh for XGBoost. Our results demonstrate that the combination of UV-Vis spectroscopy with SVC offers an efficient, rapid, and scalable alternative for Fusarium species identification, with direct applications in agricultural diagnostics. (AU) | |
| Processo FAPESP: | 21/14060-1 - Foodomics e análise multivariada no controle de micotoxinas em cevada |
| Beneficiário: | Elem Tamirys dos Santos Caramês |
| Modalidade de apoio: | Bolsas no Brasil - Pós-Doutorado |