To assess the global distribution of bread products based on sensory texture properties and instrumental measurements, PCA was performed on these data, separately. The PCA score plot of data from the DA (Fig. 2a) illustrates a huge variation between bread products based on the sensory texture attributes shown in the PCA loading plot (Fig. 2b). PC1, describing 62% of the total variation, accounted for most of the variation, representing mainly differ- ences related to type of cassava flour and addition of dietary fibers. PC2 described 18% of the total variation, representing differences in the quantity of added cassava flour (Fig. 2a). Bread made from wheat flour with added fiber (14 and 16) was characterized by a high degree of lumpiness, adhesiveness and toughness, and similar bread samples made with 30% cassava flour (15 and 17) and bread made with 50% cassava flour (5 and 10) were evaluated to be the hardest samples (Fig. 2a and b). Controls (1 and 6) and bread made with PD (12) were characterized as being drier and more breakable and negatively correlated with adhesiveness, lumpiness and toughness along PC2. Generally, bread made with Mogo (7e10) was drier and more breakable than samples made with similar quan- tities of FdM (2e5).
PCA was also performed on results from instrumental mea- surements (Fig. 2c and d). PC1 and PC2 explain 71% and 14% of the variance, respectively. PC1 was spanned by the bread products made with PD (12) and PA (13) in one end and products made with
50% cassava flour (5 and 10) and 30% cassava added 7% fiber (15 and 17) in the other end. Products made with PD (12) and PA (13) were characterized by a high volume and specific volume, a greater weight loss during baking and less pronounced as being more cohesive than the remaining samples. The samples made with the highest content of cassava flour (5 and 10) and with cassava flour and fiber (15 and 17) were harder and with a higher modulus of elasticity.
A comparison of PCA score plots based on data from the sensory and instrumental measurements (Fig. 1a and c) reveals that while PCA on sensory data separates samples in accordance to type of cassava flour (PC1) and quantity (PC2), PCA on instrumental mea- surements, only separates samples based on the type of cassava flour. Thus, TPA does not mimic the results from DA. A similar conclusion was drawn by Matos and Rosell (2012), who only found a significant correlation for a limited number of TPA parameters and sensory texture attributes in their study of gluten-free bread. In contrast, Gambero et al. (2002), who studied the textural quality of different commercial products of white pan bread, found that sensory texture attributes were well predicted by TPA. The above- mentioned studies only made pairwise comparisons between TPA results and sensory evaluation, and looking at data from a multi- variate perspective might have provided additional information.