Tumour xenograft detection through quantitative analysis of the metabolic profile of urine in mice

Abstract

The metabolic content of urine from NIH III nude mice (n = 22) was analysed before and after inoculation with human glioblastoma multiforme (GBM) cancer cells. An age- and gender-matched control population (n = 14) was also studied to identify non-tumour-related changes. Urine samples were collected daily for 6 weeks, beginning 1 week before cell injection. Metabolite concentrations were obtained via targeted profiling with Chenomx Suite 5.1, based on nuclear magnetic resonance (NMR) spectra acquired on an Oxford 800 MHz cold probe NMR spectrometer. The Wilcoxon rank sum test was used to evaluate the significance of the change in metabolite concentration between the two time points. Both the metabolite concentrations and the ratios of pairs of metabolites were studied. The complicated inter-relationships between metabolites were assessed through partial least-squares discriminant analysis (PLS-DA). Receiver operating characteristic (ROC) curves were generated for all variables and the area under the curve (AUC) calculated. The data indicate that the number of statistically significant changes in metabolite concentrations was more pronounced in the tumour-bearing population than in the control animals. This was also true of the ratios of pairs of metabolites. ROC analysis suggests that the ratios were better able to differentiate between the pre- and post-injection samples compared to the metabolite concentrations. PLS-DA models produced good separation between the populations and had the best AUC results (all models exceeded 0.937). These results demonstrate that metabolomics may be used as a screening tool for GBM cells grown in xenograft models in mice.

Global Metabolomics Reveals Urinary Biomarkers of Breast Cancer in a MCF-7 Xenograft Mouse Model

“…Previously, metabolomic studies have been carried out in human xenograft mouse models to identify urinary biomarkers of cancer using glioblastoma multiforme cells [7], gastric adenocarcinoma cells [8], and lung adenocarcinoma cells [9]. These studies were primarily undertaken using nuclear magnetic resonance spectroscopy.…”

CHAPTER 3. High‐Resolution NMR‐Based Metabolic Profiling in Drug Discovery and Development

“…Analogous model validation has been undertaken in mice bearing xenografts of human glioblastoma multiforme cells. 41 …”

A network-based feature selection approach to identify metabolic signatures in disease

“…In our experiments we use a Wilcoxon rank sum test (Moroz et al, 2011) and obtain the p-values from the test statistics. Again, we use Bonferroni correction for multiple testing.…”

“…In addition to the graph-based descriptors, we rank features using the Wilcoxon rank sum (Moroz et al, 2011) test statistics.…”

“…Filter approaches (John et al, 1994) are independent of a learning algorithm and assess features based on their ability to discriminate between predefined classes. Commonly used filter approaches are based on statistical hypothesis testing such as Student's t-test (Pan, 2002;Li et al, 2010) or Wilcoxon rank sum test (Moroz et al, 2011). Wrapper approaches (Kohavi and John, 1998) utilize a machine learning algorithm to search for a set of highly discriminating features, though generally with higher computational costs .…”