Clinical factors and liver biopsy cannot accurately predict the risk of developing cirrhosis in chronic hepatitis B (CHB). This study was to develop a predictive gene signature for cirrhosis in CHB patients. A total of 183 untreated CHB patients were enrolled. GeneChip, significant analysis of microarray (SAM) and prediction analysis of microarray (PAM) were used to select predictor genes (PGs) in liver tissues. The Cirrhosis Risk Score (CRS) was calculated based on 6 PG variables and the predictive value of CRS was evaluated. Firstly differentially expressed genes were filtered from a genome scan and SAM, and 87 significant genes were selected for the signature building. Secondly a signature consisting of 6 PGs (CD24, CXCL6, EHF, ITGBL1, LUM and SOX9) most predictive for cirrhosis risk in CHB patients was developed in the selection set (n= 40) by use of PAM and PCR approach. Finally the CRS was calculated to estimate the risk of developing cirrhosis and then tested in validation cohort (n= 143). The area under the ROC curves (AUROC) of the CRS was 0.944 and exceeded to 6 PGs and clinical factors. A low CRS cutoff of< 6.43 to identify low-risk patients would misclassify only 8.16% of high-risk patients, while a high cutoff of> 8.32 to identify high-risk patients would