A multivariate approach reveals changes in the gene expression profile associated with skeletal muscle energy metabolism in early weaned Nellore calves

Supplementary Material 1. A) Tuning of the keepX parameter in sPLS-DA for weaning gene expression data. The plot shows the balanced error rate (BER, y-axis) as a function of the number of selected variables (x-axis). The blue line represents a single component, while the red line considers up to two components; B) Tuning of the number of components in sPLS-DA applied to weaning gene expression data. Classification performance was evaluated using OER and BER for each prediction distance type: max.dist (green), centroids.dist (lilac), and mahalanobis.dist (yellow).

Supplementary Material 2. A) Tuning of the keepX parameter in sPLS-DA for slaughter gene expression data. The plot shows the balanced error rate (BER, y-axis) as a function of the number of selected variables (x-axis). The blue line represents a single component, while the red line considers up to two components. B) Tuning of the number of components in sPLS-DA for gene expression data, used to evaluate sPLS-DA classification performance (OER and BER) for each prediction distance type: max.dist (blue), centroids.dist (orange), and mahalanobis.dist (gray) using slaughter data.

Supplementary Material 3. A) Cross-validation for the selection of two components, where the y-axis represents the model predictive coefficient (q²) and the x-axis represents the number of components. B) Evaluation of variable selection in sPLS using the correlation metric. Circle size represents the mean correlation, while color indicates the standard deviation (SD). Green squares mark the optimal keepX and keepY values for each component.