Risk pathogenic genes prediction is important for uncovering the occurrence and development mechanism of complex diseases (i.e., cancer), improving the disease detection, prevention and treatment, and providing the targets for drug design. Traditional gene-mapping approaches, such as linkage analysis and genome-wide association studies (GWAS), often predict hundreds of candidate genes. But it is costly, time-consuming and laborious to further validate these candidate genes with biological experiments. However, the number of candidate pathogenic genes can be effectively reduced by computational and prioritization methods. Considering the excellent performance of random walk with restart (RWR) in predicting the risk pathogenic genes, in this work, we comprehensively discuss the recent progresses of predicting the risk pathogenic genes with RWR from databases related with genes and diseases, the metrics of measuring the similarity between genes/diseases, the strategies of choosing the seed genes of specific disease, and the different genes/diseases network structures. We also point out the computational problems and challenges faced in the process of pathogenic genes prediction.