Analysis of bee pollen constituents from different Brazilian regions: Quantification by NIR spectroscopy and PLS regression

Abstract In the present work partial least square regression (PLS) models were built for quantification of the major components of 154 Brazilian bee pollen samples. Bee pollen has nutritive and therapeutic properties that make it attractive for human health. However, studies on the nutrient and bioactive compound composition of this product are needed, as well as the verification of the presence of contaminants that are harmful to health. The conventional analysis methods are costly and time-consuming, while near infrared spectroscopy (NIR) associated to PLS regression allows a fast and non-costly quantification of the bee pollen components without samples pre-treatment. The calibration models exhibited the determination coefficients, R 2  > 0.94. The mean percent calibration error varied from 1.49 to 5.58%. For external validation, R 2 ranged from 0.89 to 0.98 among the six. The results indicated that some models are good for quantification, while others are qualified for screening calibration.