Application of the principal component analysis, cluster analysis, and partial least square regression on crossbreed Angus-Nellore bulls feedlot finished

Principal component analysis (PCA) and the non-hierarchical clustering analysis (K-means) were used to characterize the most important variables from carcass and meat quality traits of crossbred cattle. Additionally, partial least square (PLS) regression analysis was applied between the carcass measurements and meat quality traits on the classes defined by the cluster analysis. Ninety-seven non-castrated F1 Angus-Nellore bulls feedlot finished were used. After slaughter, hot carcass weight, carcass yield, cold carcass weight, carcass weight losses, pH, and backfat thickness (BFT) were measured. Subsequently, samples of the longissimus thoracis were collected to analyze shear force (SF), cooking loss (CL), meat color (L*, chroma, and hue), intramuscular fat, protein, collagen, moisture, and ashes. Principal component 1 (PC1) was correlated with colorimetric variables, while PC2 was correlated with carcass weights. Afterwards, three clusters (k = 3) were formed and projected in the gradient defined by PC1 and PC2 and allowed distinguishing groups with divergent values for collagen, protein, moisture, CL, SF, and BFT. Animals from high chroma group presented meat with more attractive colors and tenderness (SF = 1.97 to 4.84 kg). Subsequently, the PLS regression on the three chroma groups revealed a good fitness and the coefficients are used to predict the chroma variable from the explanatory variables, which may have practical importance in attempts to predict meat color from carcass and meat quality traits. Thus, PCA, K-means, and PLS regression confirmed the relationship between meat color and tenderness.