3.2 Metabolomic profiling of saline-treated vs. cisplatin-treated mice over time
XCMS processing of chromatographic data and subsequent filtering by relative standard deviation of features in pooled quality control samples resulted in the selection of 2446 features for plasma, 2319 features for urine, and 3021 features for kidney. Following pcgroup filtering (as described in the methods section), 841 plasma features, 999 urine features, and 841 kidney features remained for analysis.
PCA score plots were generated to visualize the metabolic variation between saline-treated mice and cisplatin-treated mice at each of the four timepoints. Score plots of plasma, urine, and kidney samples from C57BL/6 mice all showed a clear separation between saline and cisplatin groups from days 2-4 (Figure S1 ), with moderate separation observed at day 1 for all three sample types (Figure S1 ). Similar trends were observed in PCA score plots of FVB/N mice, with clear separation between saline and cisplatin-treated mice from day 2-4 for all sample types and weak to moderate separation at day 1 (Figure S2 ). Corresponding OPLS-DA models of pairwise comparisons between saline and cisplatin-treated mice at each timepoint confirmed the separation observed in PCA scores plots, with a high degree of fit (R2Y) and moderate predictive ability (Q2Y) between days 2-4 for both C57BL/6 (Figure 2 ) and FVB/N mice (Figure S3 ). OPLS-DA models of day 1 plasma samples for both strains of mice showed good fit and moderate predictive ability, but model statistics were poor for day 1 urine and kidney samples (Figure 2, S3 ).
In C57BL/6 mice, two-way ANOVA of individual features for each sample type revealed 181 plasma features, 683 urine features, and 66 kidney features that were significantly different by saline vs. cisplatin treatment after adjusting for multiple comparisons. Significant differences were found in 109 plasma features, 599 urine features, and 73 kidney features when comparing samples from saline and cisplatin-treated FVB/N mice. Within each sample type, significant features following similar time course patterns were grouped together by using hierarchical clustering, and clusters containing features showing early alterations were selected for further analysis (C57BL/6,Figure 3 ; FVB/N, Figure S4). To further narrow down important features, pairwise t-tests were performed between saline and cisplatin groups at each timepoint for each individual feature. Features that were significantly different in the early timepoints (day 1 or day 2) for either strain were selected for identification, summarized inTable 1 (C57BL/6) and Table 2 (FVB/N).