There are many ways to normalize biofluid metabolomics data to account for changes in dilution, all of which have been thoroughly examined in model systems. Here, urine metabolomics data was examined under relevant physiological conditions obtained from a porcine model of hemorrhagic shock and resuscitation. This includes highly variable intravascular fluid volume and urine output coupled with large perturbations in the abundance of endogenous metabolites. Seven different normalization techniques and raw data were evaluated to determine an appropriate normalization technique in this setting, including spectral post-processing methods and physiological measures of concentration. Relationships between normalization constants for each urine sample were examined, as well as relationships between urinary and serum creatinine concentrations. Principal components analysis was used to examine clustering of metabolomics data. The set of normalization constants associated with each sample were reflective of urine concentration, with a trend toward concentration decreases during late resuscitation timepoints. Urinary creatinine normalized to urine output was most reflective of serum creatinine levels. Principal components analysis showed that urine samples clustered according to experimental timepoint for all normalization methods examined. Little separation was seen in raw data. Urine output-normalized data stands out from the six other normalization methods studied because it is reflective of renal clearance and should be used when comparing urine and serum metabolomics data.
- Hemorrhagic shock