Abstract: With the completion of genome sequencing and the continuous development of gene technology, the pathogenic genes of some diseases are gradually identified. At present, people have grasped the pathogenic causes of some diseases through scientific experiments, but the pathogenic causes of most diseases, especially those related to genes, are still unknown. In this paper, the mouse data with 85% homology similarity to human is used as the research object. The disease phenotype data set, pathogenic gene data set and confirmed phenotype-gene association data set are constructed into a double-layer coupled network. The data are analyzed and mined by meta-path random walk method, and the uncertainties are predicted on the basis of confirmed phenotype-gene association data. The proposed algorithm achieves better prediction results compared with other algorithms.