Return to search

Application of ROC curve analysis to metabolomics data sets for the detection of cancer in a mouse model

The goal of this study was to show that quantifiable metabolic changes may be used to screen for cancer. NIH III nude mice (n=22) were injected with human GBM cells, with daily urine samples collected pre and post-injection. 14 mice were injected with saline to serve as controls. The measurement of metabolite concentrations took place on an 800 MHz NMR spectrometer. 34 metabolites were identified and quantified, through targeted profiling, with Chenomx Suite 5.1. Univariate statistical analysis showed that 3 metabolites (2-oxoglutarate, glucose and trimethylamine n-oxide) were significantly altered in the presence of tumour, while PCA and PLS-DA models found the maximum variance between the healthy and tumour-bearing groups. Receiver operating characteristic (ROC) curve analysis was applied to the data set to provide a measure of clinical utility. ROC statistics were as high as 0.85 for the analysis of individual metabolites, 0.939 for the analysis of metabolite pairs and 0.996 for PLS-DA models. These results show that metabolomics has potential as a screening tool for cancer. / Medical Physics

Identiferoai:union.ndltd.org:LACETR/oai:collectionscanada.gc.ca:AEU.10048/1561
Date11 1900
CreatorsMoroz, Jennifer
ContributorsGino Fallone (Oncology), Alasdair Syme (Oncology), Keith Wachowicz (Oncology), Nicola De Zanche (Oncology), Jack Tuszynski (Physics)
Source SetsLibrary and Archives Canada ETDs Repository / Centre d'archives des thèses électroniques de Bibliothèque et Archives Canada
LanguageEnglish
Detected LanguageEnglish
TypeThesis
Format1928538 bytes, application/pdf

Page generated in 0.0017 seconds