Comparative Assessment of WINISI and The Unscrambler for Near Infrared Reflectance Spectroscopy (NIRS) Modelling of Phytate and Antioxidants in Brown Rice

Abstract

Abstract Brown rice has been known to be a better alternative to polished rice due to greater amounts of nutrients including antioxidants. Phytic acid and phenols in brown rice are also known to impart antioxidant capacity. Near infrared reflectance spectroscopy is renowned to estimate the nutritional composition of brown rice. The present study was conducted to develop the prediction models of total phenolic content (TPC), phytic acid (TPA) and antioxidant capacity (TAC) of brown rice. While both WIN ISI and The Unscrambler® software produce crisp models, certain limitations such as principal component analysis (PCA) descriptives in WIN ISI and inverse multiple scatter correction (iMSC) in The Unscrambler® lead to uncertainty of the results. Hence, the models in this study were generated using PCA and partial least square (PLS) regression, which were compared on both WIN ISI and The Unscrambler® separately. The reference data of 226 rice landraces was subjected to both the softwares and optimal models were obtained from standard normal variate (SNV-DT) for TPA and TPC over The Unscrambler®, while multiplicative scatter correction (MSC) was found better for TPC model development on WIN ISI. According to regression analysis the best prediction model was obtained for TPC employing MSC with RSQ = 0.925 and RPD = 3.11 in WIN ISI, while the rest were better validated in The Unscrambler® using SNV-DT with RSQ = 0.888, 0.958 RPD = 2.97, 2.93 for TPA and TAC respectively. The results are indicative of the NIRS ability and the effect of different scatter corrections to rapidly predict antioxidant content in brown rice.