PDZ domains are one of the most well studied peptide binding domains. These domains usually bind short peptides at the C-terminus of their target proteins and play a crucial role in cellular signalling processes. Computational approaches have been published to determine the interaction specificity of PDZ domains, but these prediction methods often limit their predictions on a limited subset of PDZ domains. In this research work, we developed PreDiZ, a computational method for PDZ domain-peptide interaction prediction based on the SDR approach. The SDR approach was originally created to predict specificity of protein kinases. In this work, improvements have been made to apply the SDR approach to the PDZ domains, including using a more sophisticated strategy to determine SDRs, and using both positive and negative interactions in the prediction. As a result, PreDiZ is able to work on a wide range of PDZ domains, including novel PDZ domains. In cross-validations, PreDiZ scored AUCs range from 0.82 to 0.94. In the comparison against published methods, PreDiZ showed competitive performance on making prediction to distantly related PDZ domains, but not as good as other recently published methods on mouse test set.However, the results also suggested that PreDiZ could be improved by optimising SDRs. We also conducted proteome-wide predictions on A. thaliana, C. elegans, D. rerio, M. musculus and H. sapiens and showed PDZ domains were evolved relatively late in eukaryotic cells. Network studies on human PDZ domain interaction revealed the enriched GO terms and KEGG pathways of PDZ domain binding proteins. Lastly, we studied how H7N9's NA and H5N1's NS1 protein regulate human biological processes using the human PDZ interaction network. Human proteins that regulated via PDZ domain interactions by these two kind of viral proteins, were enriched in similar biological processes. Therefore, we concluded that the function of PBM in the NS1 proteins of H5N1 was replaced by the PBM in the NA proteins of H7N9.