10 environment (Math-Works, Natick, USA) with the PLS-toolbox version 4.0 (Eigenvector Research, Inc., Wenatchee, WA). Samples

(n = 139) were divided into calibration (n = 100) and validation sets (n = 39) by applying the classic Kennard–Stone (KS) selection algorithm to the NIR spectra ( Kennard & Stone, 1969). Calibration and validation sample numbers were, respectively selected at 100 (min. 1.50 g kg−1; max. 82.0 g kg−1; mean 28.6 g kg−1; SD 20.8 g kg−1) and 39 (min. 2.20 g kg−1; max. 68.5 g kg−1; mean 15.8 g kg−1; SD 14.8 g kg−1). Outlier detection was executed to improve model accuracy, remove samples with extreme values, which exhibit increased influence on the model, and eliminate unmodelled residues in the X and Y data responses.

An elliptical joint confidence region (EJCR) was GSK1120212 datasheet calculated to evaluate the ERK inhibitor slope and intercept for the reference regression, and predict values at a 95% confidence interval. Mathematics and applications for the EJCR can be addressed in the following references: González, Herrador, and Asuero (1999), and Goicoechea and Olivieri (2001). All calculations were performed using suitable MATLAB 7.10 approaches. Açaí and palmitero-juçara sample spectra (sample #1 and #80), with respective 3.3 and 36.9 g kg−1 total anthocyanin content (TAC) are presented in Fig. 1. Fig. 1 depicts two spectra, (a) and (b), which characterise the following: a second overtone of the O–H stretch at 982 nm, the first overtone of the O–H stretch

at 1456 nm, and combination bands of the asymmetric and scissor stretch O–H vibrations at 1940 nm. Sugar-related absorption bands were also observed at 926, 1208, 1728, 1762, 2308, and 2348 nm. The spectra were quite homogeneous, and no outliers were identified by a priori visual inspection. The PLS method was initially performed on the entire original spectra wavelength range to develop the NIR model. In this way, TAC could be predicted non-destructively. Noise and systematic behaviour were undesirable features in the spectra, and therefore pre-processing Tacrolimus (FK506) was necessary. The original spectra were subjected to smoothing (first order), multiplicative scattering correction (MSC), and first, second order derivatisation (Savitzky–Golay). Calibration model results for the TAC NIR region in açaí and palmitero-juçara fruits are shown in Table 1. In addition to the PLS models, PLS-SPA, PLS-GA, and iPLS model results are shown. The best performing model results from pre-processing tests (see methods) are provided. F-tests were performed for all models using the prediction set. The results detected no significant differences (95% confidence level) between the best PLS model, the PLS (4) and other models, with the exception of PLS (4) smoothing (3 pts.), PLS (4) first derivative (7 pts.), PLS (4) second derivative (3 pts.