Discovery of novel metabolite biomarker(s) for improved ovarian cancer diagnosis is of great importance. In this paper, the differences of urinary hydrophilic and hydrophobic metabolic profiling between healthy women, benign ovarian tumor, and ovarian cancer patients were studied by metabolomics strategy. Metabolites in urine were analyzed on hydrophilic interaction chromatography (HILIC) and reversed-phase liquid chromatography (RPLC) coupled to MS. Data from HILIC or RPLC, positive or negative ion detection mode were found to be complementary. Data were filtered by orthogonal signal correction (OSC) method, and the three groups were discriminated by partial least squares discrimination analysis (PLS-DA) models. By combining the four datasets, maximum information can be collected, and a PLS-DA model was built after OSC filtering. The model based on combined dataset is superior to the ones based on the separate dataset, and important metabolites were screened based on the combined dataset model. Five metabolites were found to be specific to ovarian cancer and ten metabolites were considered commonly related to ovarian cancer and benign ovarian tumor. Combination of RPLC and HILIC separation, as well as positive and negative ion detection in metabolomic studies show advantages in collecting various metabolites information that helps us better understand the metabolic event.