Simca p v13

Statistical analyses were performed using SIMCA P+ v13.0 (UMETRICS) and R (R-Project for Statistical Computing, CRAN.R-project.org). PCA was performed on all samples together as well as individual analyses for each class. This separate analysis was used to remove any outlying samples that were not highly similar to the remaining members of that class. Only a single sample was removed by this process, for details see the supplementary material. After this trimming process PCA and OPLS-DA was performed on the entire dataset. OPLS-DA models were then generated for each combination of control and treated samples (e.g., Bp control and Bp Al exposed, Sc control spent media and Sc Cu exposed spent media etc.) in order to minimize the amount of variation being examined in any one model, thereby maximizing the interpretability of each model. For each of these pairwise OPLS-DA models the R₂Y, Q₂ and CV-ANOVA p values were used to assess model quality; only models with CV-ANOVA p values below 0.05 were accepted as statistically significant. From each significant model the VIP and p(corr) values were exported for further interpretation. Shared and unique structures plots were used as this type of plot simplifies analysis of metabolomics data while maintaining the depth of complexity within the dataset (Wiklund et al., 2008 (link)).

Fecal samples were lyophilized, derivatized and analyzed by gas chromatography time-of-flight mass spectrometry (GC-TOF-MS) (Agilent 7890 gas chromatograph and LECO Pegasus III time-of-flight mass spectrometer) as previously described [23 (link),24 (link)]. MS-DIAL software [25 (link)] and FiehnBin base database were used for raw peaks exacting, the data baselines filtering, and calibration of the baseline [26 (link)]. Peaks detected in ≤50% of QC samples or <50% samples of every group were removed, except QC group or RSD>30% in QC samples [27 (link)]. SIMCA-P v13.0 (Umetrics, Umea, Sweden) was used for partial least squares-discriminant analysis (PLS-DA) and orthogonal projections to latent structures-discriminant analysis (OPLS-DA). The first principal component of variable in importance projection (VIP) was obtained to refine the analysis. VIP > 1.5 was first selected as “changed metabolites”. Obtained metabolites were validated by searching in the Kyoto Encyclopedia of Genes and Genomes (KEGG) [28 (link)].

Differences between males and females in the study were tested with Mann-Whitney U-test using Statistica (StatSoft, Oklahoma, USA) since most of these variables did not show normal distribution. Repeated measurement analysis of variance (ANOVA) was used to analyses differences in saliva collection with Bonferroni test as post hoc test when the ANOVA indicated significant differences.
Correlations between variables were tested for Statistical significance with Spearman correlation test, adjusted for multiple comparisons according to Bonferroni. Descriptive data are presented as mean and standard deviation (SD) or median and interquartile range (IQR). For all analyses, the significance level was set at P < 0.05.
When investigating the multivariate correlations between the proteins and group membership. Orthogonal Partial least squares discriminant analysis (OPLS-DA) was applied using SIMCA-P+ v.13.0 (UMETRICS, Umeå, Sweden)8 (link)32 33 (link). Principal component analysis (PCA), also available in SIMCA-P+ was used to identify multivariate outliers. No multivariate outliers were identified. The variable influence on projection (VIP) indicates the relevance of each X-variable pooled over all dimensions and the Y-variables – the group of variables that best explains Y. VIP > 1.2 was considered significant.