Quantification of quality parameters in castanhola fruits by NIRS for the development of prediction models using PLS and variable selection algorithms on a laboratory scale

Abstract

This paper proposes a novel methodology for the quantification of total phenolic compounds (TPCs) and total anthocyanin compounds (TACs) in castanhola fruits (Terminalia catappa Linn), using near infrared spectroscopy (NIRS) coupled with variable selection algorithms, such as interval partial least squares (iPLS) and genetic algorithm-partial least squares (GA-PLS). GA-PLS showed the best results in the prediction of both parameters. TPC parameters, R_p² = 0.82, RMSEP = 11.3 mg GAE g⁻¹ (mg gallic acid equivalents (GAE) per g sample), SEL 16.70 mg GAE g⁻¹, RPD = 2.89, sensibility 2.19 × 10⁻⁷, and selectivity 0.048 were also obtained using first derivative (5 points) and MSC pretreatment. TAC parameters, R_p² = 0.80, RMSEP = 8.70 mg L⁻¹, SEL 6.93 mg L⁻¹, RPD = 1.90, sensibility 6.73 × 10⁻⁶ and selectivity 0.07 were attained using second derivative (11 points) pre-treatment. From these findings, it can be concluded that NIRS coupled GA-PLS can be used as a non-destructive technique for determining TACs and TPCs in intact castanhola fruits.