Abstract： Objective To characterize the metabolic "fingerprint" of fecal extracts for diagnosis of early-stage colorectal cancer (CRC) using proton nuclear magnetic resonance spectroscopy (1H-NMR)-based metabolomics coupled with pattern recognition.Methods From January 2014 to December 2014,we collected fecal samples at the Second Affiliated Hospital of Shantou University Medical College,from 25 patients with colorectal adenomas (CR-Ad),20 with stage Ⅰ /Ⅱ CRC,and 32 healthy controls (HCs).The patients were diagnosed by histopathology.No subjects had any complicating diseases.HCs showed no abnormalities from blood tests,endoscopic examination,diagnostic imaging,and/or medical interviews.We excluded participants who used antibiotics,NSAIDS,statins,or probiotics within two months of study participation,and any patients who underwent chemotherapy or radiation treatments prior to surgery.We used orthogonal partial least-squares-discriminant analysis (OPLS-DA) for pattern recognition (dimension reduction) on 1H-NMR processed data (1H frequency of 400.13 MHz),to find metabolic differences among CR-Ad,carcinoma and HC fecal samples;and receiver operating characteristic (ROC) analysis to determine the diagnostic value of the fecal metabolic biomarkers.Results Fecal samples were collected from 20 patients with Stage Ⅰ/Ⅱ CRC (11 M,9 F,median age (52±13) years),25 with CR-Ad (14 M,11 F,median age (53±11) years) and 32 HCs (15 M,17 F,median age (53± 14) years).OPLS-DA clearly distinguished CR-Ad and stage Ⅰ/Ⅱ CRC from HC samples,based on their metabolomic profiles.Relative signal intensities in HCs were significantly lower than in the cancer patients for butyrate (HC:23.0±6.0;CR-Ad:18.0±5.0;CRC:14.0±6.0;Z=-2.07,P=0.008),acetate (HC:45.0± 11.0;CR-Ad:31.0±11.0;CRC:24.0±8.0;Z=-2.32,P=0.011),propionate (HC:26.0 ± 7.0;CR-Ad:22.0 ± 6.0;CRC:19.0 ± 5.0;Z=-2.43,P=0.032),glucose (HC:37.0±7.0;CR-Ad:31.0±7.0;CRC:26.0±8.0;Z=-2.07,P=0.044) and glutamine (HC:4.5±2.0;CR-Ad:4.9 ± 1.0;CRC:5.4 ± 1.0;Z=2.21,P=0.044).However,relative signal intensities in HCs were significantly higher than in patients for lactate (HC:4.8± 1.0;CR-Ad:6.9±2.0;CRC:4.8± 1.0;Z=2.02,P=0.038),glutamate (HC:3.2±2.0;CR-Ad:4.9 ± 1.0;CRC:3.2± 2.0;Z=2.21,P=0.044) and succinate (HC:12.0±2.0;CR-Ad:15.0±3.0;CRC:12.0± 2.0;Z=2.25,P=0.011).Among the potential biomarkers,acetate at 1.92 ppm,and succinate at 2.41 ppm displayed relatively high area under ROC,with sensitivity and specificity both ＞90％,to distinguish early-stage CRC patients from HCs.Conclusion Fecal metabolic profiles distinguish of HCs from patients with CRC patients,even in the early stages (stage Ⅰ/Ⅱ),highlighting the potential of NMR-based fecal metabolomic fingerprinting as tools for early CRC diagnosis.