Discovery of disease biomarker based on untargeted metabolomics is informative for pathological mechanism studies and facilitates disease early diagnosis. Numerous of metabolomic strategies emerge due to different sample properties or experimental purposes, thus, methodological evaluation before sample analysis is essential and necessary. In this study, sample preparation, data processing procedure and metabolite identification strategy were assessed aiming at the discovery of biomarker of breast cancer. First, metabolite extraction by different solvents, as well as the necessity of vacuum-dried and re-dissolution, was investigated. The extraction efficiency was assessed based on the number of eligible components (components with MS/MS data acquired), which was more reasonable for metabolite identification. In addition, a simplified data processing procedure was proposed involving the OPLS-DA, primary screening for eligible components, and secondary screening with constraints including VIP, fold change and p value. Such procedure ensured that only differential candidates were subjected to data interpretation, which greatly reduced the data volume for database search and improved analysis efficiency. Furthermore, metabolite identification and annotation confidence were enhanced by comprehensive consideration of mass and MS/MS errors, isotope similarity, fragmentation match, and biological source confirmation. On this basis, the optimized strategy was applied for the analysis of serum samples of breast cancer, according to which the discovery of differential metabolites highly encouraged the independent biomarkers/indicators used for disease diagnosis and chemotherapy evaluation clinically. Therefore, the optimized strategy simplified the process of differential metabolite exploration, which laid a foundation for biomarker discovery and studies of disease mechanism.