Construção de modelo para predição proteica do tegumento de feijão carioca por meio do Near Infrared Spectroscopy (NIRS)

Abstract

Bean (Phaseolus vulgaris L.) has a great nutritional, cultural and food safety importance, being a traditional Brazilian food. Consumers prefer the carioca beans grains with light background tegument, and reject the grains when presenting a dark background. After the harvest, tegument darkening occurs due to the enzymes action, such as polyphenoloxidases and peroxidases. A classical method for total proteins quantification is Bradford, which consists of the protein-dye binding, representing a fast, stable and destructive procedure. Near Infrared Spectroscopy (NIRS) is a rapid, non-chemical and non-destructive methodology. The study general objective was the construction of a protein prediction model for carioca beans with two distinct granulometries: bran and flour, using NIRS. To obtain the grains, the field experiment was carried out in Pato Branco - PR municipality, in UTFPR experimental area, using 10 carioca beans genotypes: BRS MG Madrepérola; Perola; TAA Dama; IAC Milênio; ANFC 9; IPR Tangará; BRS Requinte; IPR Curió; BRS Imperador; BRS Estilo. At physiological maturity stage, grains were stored in transparent plastic trays, in a controlled chamber under 40 ± 5 ºC temperature and 75% humidity for until four months, being harvested at 0, 15, 30, 60, 90 and 120 days. NIRS requires reference values to develop the calibration model. In this case, Bradford (1976) methodology was used to quantify the total proteins. Statistical tests consisted in analysis of variance followed by Tukey test at 95% significance. NIRS sample preparation implicated in whole grain crushing until to obtain the flour and bran. In the spectrometer, 20 g of samples were accommodated in a rotating quartz cuvette. Two methods of prediction were used: Cross Validation and Test Set. The methods were compared and it was evaluated the best protein contents prediction. Observing separately the Bradford results, analysis of variance indicated significant interaction between time of storage x cultivar. At 60 and 90 days of storage, the highest variability occurred among the cultivars. These times were the ones wich better expressed the potential for tegument darkening, evaluated by the protein content. The cultivars IAC Milênio and ANFC 9 did not differ statistically in any time of storage, the tegument protein content remained constant and the cultivars were considered stable. The limits used to generate PLS models were: first quartile at 16.78 mg g-1 and third quartile at 17.63 mg g-1. The flour bean granulometry was the most efficient for construction of the prediction algorithms, since possibly generated more homogeneous spectra. Among the flour group, the prediction model that presented the best indexes was Test Set. The model is not finished due to intrinsic limitation in the protein variability of the studied cultivars.