FIG. 6.
Multivariate data analyses. (A) PLS-DA score plot containing two first components. Each plot mark corresponds to an observation (individual rat spinal cord sample). The confidence ellipses are based on the Hotelling T2 and illustrate the 95% confidence regions. Principal component 1 (PC1; x-axis) describes differences in the spectra between sham and injury operations. Principal component 2 (PC2; y-axis) shows a marked separation between control and ω-3 diets. Score plot analyses revealed statistically significant separations between groups (p<0.01 by permutation test). (B) Loading plot between PC1 and PC2. Loading 1 explains differences between sham and injured animals, whereas features in loading 2 explain differences between dietary interventions. Groups with stronger impact are further away from the plot origin (represented by dashed lines). Data demonstrate that ω-3 polyunsaturated fatty acids (PUFAs) contributed significantly to lipidomic differences between injured animals (location in loading 2 is far from zero). In contrast, analyses revealed that ω-6 PUFAs present a major contribution to both model loadings (loading 1: surgery effects and loading 2: diet effects), validating their role in chronic spinal cord injury and modulation by dietary ω-3 PUFAs. Correlated metabolites were grouped together to facilitate data presentation. Color image is available online at www.liebertpub.com/neu