Serum Metabolic Fingerprinting to Detect Human Nasopharyngeal Carcinoma Based on Gas Chromatography-Mass Spectrometry and Partial Least Squares-Linear Discriminant Analysis

Abstract

In this paper, metabolic fingerprints were obtained from 102 nasopharyngeal carcinoma (NPC) patients and 107 healthy adults by gas chromatography-mass spectrometry (GC/MS). Partial least squares-discriminant analysis (PLS-DA) has revealed a pattern recognition discriminating the patients from controls, which sensitivity is 89.72% (96/107) and specificity is 85.29% (87/102). Furthermore, double-blind experiment was carried out and satisfactory results were obtained (total correct rate 87.30%). In addition, metabolites that most strongly influence this separation were obtained. The results indicated that a metabonomic approach is feasible and efficient and deserves further evaluation as a potential novel strategy for the detection of nasopharyngeal carcinoma.

Keywords:

• Metabonomics
• Nasopharyngeal carcinoma
• PLS-DA

Acknowledgments

This work is financially supported by the international cooperation project on traditional Chinese Medicines of Ministry of Science and Technology of China (No. 2007DFA40680), National Nature Foundation Committee of China (Grant No. 20875104), China Postdoctoral Science Foundation funded project (No. 20080440181), Special Fundation of China Postdoctoral Science (No. 200902481), and Central South University Science Development Foundation (No. 10SDF22).

Notes

Note: There are no differences between NPC patients and healthy controls as regards to diabetes mellitus, cardiovascular disease, hepatic/renal disease, and blood system disease.

^a UICC: International Union Against Cancer.

^a Sensitivity is the number of true positives classified as positive. Specificity is the number of true negative classified as negative. Predictive ability of the multivariate models was evaluated by 10-fold cross validation.

Note: data are presented as means ± SD. t: t-test results between NPC patients and controls; A p value of <0.05 was considered statistically significant and signed t value is “1,” otherwise “0.”

Lun Zhao Yi and Dan Juan Li contributed equally to this work.