Application of L-EDA in metabonomics data handling: global metabolite profiling and potential biomarker discovery of epithelial ovarian cancer prognosis

Abstract

Solution capacity limited estimation of distribution algorithm (L-EDA) is proposed and applied to ovarian cancer prognosis biomarker discovery to expatiate on its potential in metabonomics studies. Sera from healthy women, epithelial ovarian cancer (EOC), recurrent EOC and non-recurrent EOC patients were analyzed by liquid chromatography-mass spectrometry. The metabolite data were processed by L-EDA to discover potential EOC prognosis biomarkers. After L-EDA filtration, 78 out of 714 variables were selected, and the relationships among four groups were visualized by principle component analysis, it was observed that with the L-EDA filtered variables, non-recurrent EOC and recurrent EOC groups could be separated, which was not possible with the initial data. Five metabolites (six variables) with P < 0.05 in Wilcoxon test were discovered as potential EOC prognosis biomarkers, and their classification accuracy rates were 86.9% for recurrent EOC and non-recurrent EOC, and 88.7% for healthy + non-recurrent EOC and EOC + recurrent EOC. The results show that L-EDA is a powerful tool for potential biomarker discovery in metabonomics study.