2.7 Statistical Analysis
The EZInfo 3.0 software (Umetrics, Umeå, Sweden) was used to perform
multivariate analysis on the metabolomics dataset. Data was centered,
pareto scaled and subsequently analyzed by principal component analysis
(PCA), an unsupervised approach to visualize the metabolic differences
between no AKI and AKI patients at each timepoint. Orthogonal partial
least squares discriminant analysis (OPLS-DA), a supervised
discriminatory analysis, was used for the pairwise discrimination of no
AKI and AKI patients at each timepoint. For each OPLS-DA, metabolites
were ranked by their variable importance in projection (VIP) values and
features with VIP values ≥ 1 were considered to have discriminatory
value in discriminating between no AKI and AKI. This VIP filtering was
repeated until OPLS-DA model statistics (R2 and Q2 values) were
maximized to select for the most important features to annotate. The
final optimized OPLS-DA model was used to generate a list of features to
identify, using a VIP value threshold of ≥ 1 and correlation (p(corr))
values less than -0.4 and greater than 0.4.
Features determined to have discriminatory value were analyzed by
two-way ANOVA with Benjamini-Hochberg false discovery rate (FDR)
correction. Individual features that were found to be significantly
different by AKI classification following two-way ANOVA and FDR
correction were further analyzed by pairwise t-tests comparing no AKI
and AKI patients at each timepoint, with p-values adjusted for multiple
comparisons using Bonferroni correction. To find metabolites altered
over time, serum and urine features were analyzed by one-way ANOVA with
FDR correction, followed by Tukey’s test for metabolites significant by
one-way ANOVA after FDR correction. p<0.05 was considered as
significantly significant for all univariate data analysis.
Receiver operating characteristic (ROC) curves were generated and area
under the ROC (AUROC) values were calculated using MetaboAnalyst 5.0.