Desenvolvimento de modelos de calibração multivariada para determinação de compostos fenólicos individuais em própolis

Abstract

Propolis is a resinous material, used for various purposes in the hive. Currently, research on the use of propolis bioactive properties has increased, thus showing the importance of propolis, and how rich its chemical composition is. The present work has as objective, to elaborate models of multivariate calibration, using the spectra obtained by the technique of NIRS, as well as, the quantifications obtained by the technique of CLAE, evaluating the figures of merit obtained. In addition to determining the use of chemometry associated with NIRS. The models for trans-cinnamic acid, galangin and cumaric acid were built using the MatLab program, a PLS package. Other tools such as HCA and PCA, with the ChemoStat program, were used to remove outliers and apply the best pre-processing. For the three models, the RER was above 10, showing high practical utility, whereas the R² calibration coefficient was between 0.90 and 0.96, the prediction R² between 0.84 to 0.90. Among the three mode is, the one with the best values for the figures of merit, was the model of trans-cinnamic acid, with the following errors: RMSEC of 0.224 mg g-¹, RMSECV of 0.437 mg g-¹ and RMSEP of 0.476 mg g-¹. The following determination coefficients were obtained: R² calibration of 0.91, R² cross-validation of 0.67 and R² prediction of 0.84, the values obtained for the coefficients were consistent with those found in the literature. The three models built presented figures of merit considered reproducible, so it is possible to conclude that NIRS can be used together with chemometry to predict individual phenolic compounds in propolis, through pre-processing and elimination of outliers.