Abstract: The identification of disease genes is the crucial step in uncovering disease pathology and systematically analyzing polygenetic disease. The high-throughput technology has advanced the development of network-based approaches for disease gene prediction. Based on the "guilt-by-association" principle, now disease gene prioritization methods can measure the proximity between candidate genes and causal genes so as to pinpoint the potential disease genes. In this review, we first classify the network-based approaches for disease gene prediction into three categories:the approach based on disease genes information, the approach integrated with phenotype similarity and the approach that integrates several results from multiple data resources into one final result. Then we bring out the current situation of these approaches and summarize the current achievements and existing problems. Finally we put forward some suggestions for future research.