To corroborate this argument, the model was also validated by means of a permutation analysis involving 1000 permutations of the class label and the computation of the accuracy of the PLS-DA model with 7 components (as indicated in the LOOCV-based analysis) in each one. The relative position of the accuracy of the model in relation to the original class is clearly an outlier in com- parison with the ones of the permutated sample, leading to a p- value less than 0.1% (Fig. 5(b)). The p < 0.001 implies that the ac- curacy of a permuted PLS-DA model would rarely be better than the one of the model adjusted to the original data.