For example, considering the retention index and mass spectrum, peak #18 corresponds to an 8-carbon aliphatic acid, but not exact identification could be attributed. Also in seven chromatographic BKM120 chemical structure peaks no identification was possible. Since these unidentified or partially identified peaks could be related to the sensorial properties of the samples and, therefore, could be significant to the PLS models, it was retained on the input data. Most of these compounds had already been identified in previous studies on the composition of the volatile
fraction of Pilsner beers and can be related to the brewing process. After GA variable selection, 11 variables were selected for the bitterness parameter (Table 1). This corresponds to a reduction of approximately 80% of the 54 original variables. Also in Table 1, the selected peaks by OPS method to the bitterness parameter are presented. Here, it was pointed out 17 variables, representing a reduction of approximately this website 68.5% of the original variables. In the OPS selection, it was evaluated different informative
vectors and combinations of vectors such as the regression (R), the root square (S) error, the net analyte signal (NAS) vectors, and combinations of NAS and S (NS) vectors and R and S (RS) vectors. Comparing the results from all of them evaluating the RMSECV and the correlation coefficients of the obtained models, the best result was obtained utilizing the NS combination
vector. From the selected peaks by the GA and OPS approaches, seven were pointed out commonly. It corresponds to approximately 64% of agreement in the selection performed by OPS relating to the one carried out by the GA. The Table 2 presents some parameters of the best models to the GA and OPS selection methods. Considering the selected peaks commonly pointed out by both approaches, the compounds probably closed related with the bitterness attribute are ethyl acetate, 1-octanol, p-vinylguaiacol, γ-nonalactone, β-phenylethyl butyrate, caryophyllene oxide and dibutylphthalate. Using only these Carbohydrate selected variables, it is possible to study the bitterness attribute, since really relevant information was captured. This means that the selected variables are the ones that are directly related to the bitterness quality parameter. It is important to emphasise that in most models obtained by OPS method utilizing other informative vectors, even among other variables, the compounds cited above were always selected. Ethyl acetate (#3 in Table 1) is an ester derived from ethanol and acetic acid and, as with most esters; it is correlated with the freshness and fruitiness of young beers (Wampler, Washall, & Matheson, 1996). 1-Octanol (#14 in Table 1) has also already been reported in the volatile fraction of beer (Pinho et al., 2006).