Prediction of sugars in ripe banana pulp from spectra of unripe fruit peel using near-infrared spectroscopy and multivariate calibration

• Ano de defesa: 2024
• Autor(a) principal: Abreu, Maria Luísa Soares de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Viçosa; Agroquímica
• Palavras-chave em Português: Açúcares (Química Orgânica); Banana; Espectroscopia de infravermelho próximo; Quimiometria; Cromatografia líquida de alta pressão; Sacarose; Glicose; Frutose; CIENCIAS EXATAS E DA TERRA::QUIMICA::QUIMICA ANALITICA
• Link de acesso: https://locus.ufv.br/handle/123456789/32697; https://doi.org/10.47328/ufvbbt.2024.251

The prediction of sucrose, glucose, and fructose concentrations in ripe banana pulp from spectra of both unripe and ripe fruit peel employing near-infrared spectroscopy (NIR) and multivariate calibration is the aim of this work. 320 unripe bananas were harvested, and NIR spectra were collected on the peel using diffuse reflectance. The sugars, i.e., sucrose, glucose, and fructose, were extracted from the pulp using an adapted procedure presented in this work. Subsequently, the extract was analyzed by high-performance liquid chromatography (HPLC) with an evaporative light-scattering detector. The sucrose, glucose, and fructose concentrations ranged from 45.10-207.20, 23.81-78.18, and 21.60-85.26 mg/g, respectively. Multivariate calibration models using partial least squares regression (PLS) were built to predict sugars from NIR spectra. Ordered predictors selection (OPS) was employed to select the most informative variables. The PLS models from the unripe peel for sucrose, glucose, and fructose presented root mean square error of prediction (RMSEP) values of 15.18, 3.41, and 8.7 mg/g, respectively. For the dataset using the ripe banana peel, the RMSEP values were 19.29, 5.29, and 6.65 mg/mL, respectively. Keywords: Banana; NIR; Chemometrics; PLS; HPLC; Sucrose; Glucose; Fructose; Early prediction