Could antioxidant capacity and flavonoid content of ethanolic extracts of geopropolis from Brazilian native bees be estimated from digital photos and NIR Spectra?

Abstract

In the current study, Multi Imaging Analysis (MIA) of digital photographs and NIR spectra of ethanolic extracts of geopropolis (EEGs) from Brazilian native bees were used for the first time to build partial least square regression (PLS) models to predict the antioxidant capacity (AC) and Total Flavonoid Content (TFC) of EEGs. Firstly, spectrophotometric reference assays were carried out: TFC was determined by complexation with AlCl3 and AC by the antiradical scavenging of the stable radical 2,2-diphenyl-1-picrylhydrazine (DPPH), Ferric Reducing Antioxidant Power (FRAP) and Cupric Ion Reducing Antioxidant Capacity (CUPRAC). Subsequently, NIR spectra and digital photos of EEGs were taken and the raw data was calibrated using PLS regression. All Color-based and NIR-based PLS models were appropriate to estimate AC and TFC in EEGs but the best performance and precision were obtained using Color-based PLS models (R2 for calibration ≥0.74 together with low RMSEC and RMSECV and lesser latent variables). The predictive capacity of Color-based and NIR-based PLS models was similar, although MIA-TFC and MIA-DPPH models indicated more reliability and accurate predictive capacity (R2 > 0.80 together with low RMSEP). Consequently, simple digital photos of geopropolis extracts or NIR spectra combined with PLS models can be potentially used to predict the total flavonoid content and antioxidant capacity of geopropolis with acceptable accuracy. This approach might be employed in the control quality of geopropolis extracts with the advantages of being a fast, low cost and green methodology.