Knowledge of the adaptation mechanism of enzymes to extreme pH values and distinguishing them from one another are necessary in the proteomics field, and would help in the drug design of stable enzymes. In this work, we have systematically analyzed the information of 105 acidic and 112 alkaline enzymes, and propose an approach for distinguishing acidic enzymes from alkaline enzymes by combining the amino acid composition, reduced amino acid composition, gene ontology, evolutionary information, and auto covariance of averaged chemical shift (acACS). The overall prediction accuracy is 94.01% by 10-fold cross-validation using the algorithm of support vector machine. This result is better than that obtained by other existing methods. The improvement of the overall prediction accuracy reaches up to 3.3% higher than those of the random forest algorithm and secondary structure amino acid composition. The acACS performance is excellent, indicating that our approach is better than other existing methods in the literature. A user-friendly web-server pred-enzymes for predicting acidic and alkaline enzymes has been established, which is aceessible to the public. (C) 2013 Elsevier Ltd. All rights reserved.