Objective:As a highly lethal disease,esophageal squamous cell carcinoma(ESCC) shows a relatively high prevalence in China. Since patients with early-stage tumors usually showed no symptoms or signs and there were no general screening methods,ESCC was often diagnosed at late stages,which leading to poor prognosis. Therefore,we performed a metabolomics profiling approach to discover noninvasive and reliable serum biomarkers of ESCC for rapid-screening diagnosis among Chinese Han population. Methods:The proton nuclear magnetic resonance (1H NMR) spectra of 29 ESCC patients and 31 age- and sex-matched healthy controls were firstly analyzed using partial least squares-discriminant analysis (PLS-DA),neural networks analyses and hierarchical cluster analysis (HCA). The serum metabolomic fingerprint discovered was then validated in additional 20 ESCC cases and 21 controls from another medical center. Results:The metabolomic PLS-DA score plot showed good separation between ESCC and healthy groups. The following neural networks analyses and HCA identified an ESCC-related metabolomites combination including two markers (serum praline and glutamine/glutamate). Conclusion:Our results highlight the NMR-based metabolomics approach in distinguishing individuals with or without ESCC and support the potential application as early diagnosis of ESCC.