| Colon cancer is the third most common cancer in people, and early diagnosis is critical to survival. This study investigates the efficacy of using metabolomic technology in diagnosing colon cancer in a mouse model. Urine from a genetically defined population of mice was analyzed by NMR spectrometry, after carcinogen exposure and categorization for tumor development based on histological examination. The NMR spectra were then analyzed by statistical methods of classification to determine if colon tumors result in changes to the metabolites secreted in urine that can be detected by NMR spectrometry. Different statistical analyses were also compared to determine which is most effective at retrieving information from the NMR data. Principal Component Analysis (PCA), a popular analysis for metabolomic data, is ineffective on this data set. Support Vector Machine (SVM) reveals six significant components, which when entered into PCA results in clear separation of normal and tumor-bearing mice. |
