1H nuclear magnetic resonance (NMR)-based metabonomics was used to characterize metabolic profiles of liver cirrhosis (LC) and hepatocellular carcinoma (HCC). We found compared to healthy humans, LC and HCC sera had higher levels of acetate, Nacetylglycoproteins, pyruvate, glutamine, alpha-ketoglutarate, glycerol, tyrosine, 1-methylhistidine and phenylalanine, together with lower levels of low-density lipoprotein, isoleucine, valine, acetoacetate, creatine, choline and unsaturated lipids. Scores plot of pattern recognition analysis were capable of distinguishing LC and HCC patients from healthy humans. These results indicate that serum NMR spectra combined with pattern recognition analysis techniques offer an efficient, convenient way of depicting tumor biochemistry, which may be of great benefit to early diagnosis of human malignant diseases using single blood samples.