In　the　existing　genetic　similarity　search　algorithm　based　on　meta-path,　the　accuracy　of　genetic　similarity　calculation　results　is　low　because　the　implicit　correlation　between　genes,　diseases　and　other　related　factors　is　not　taken　into　account.　To　solve　this　problem,　an　improved　weighted　meta-path　genetic　similarity　search　algorithm　gSim-Search　is　proposed.　This　algorithm　uses　binary　network　to　spread　resources.　It　not　only　reconstructs　the　relationship　between　nodes　in　gene-disease-phenotype　heterogeneous　networks,　but　also　assigns　reasonable　weights　to　the　relationship,　to　express　the　degree　of　correlation　of　nodes　and　reflect　the　similarity　of　genes　scientifically.　It　solves　the　problem　of　sparse　connection　and　insufficient　information　in　traditional　metapath-based　methods.　Experiments　show　that　the　algorithm　greatly　improves　the　accuracy　of　predicting　genetic　similarity　between　breast　cancer　and　obesity.　©　2020,　Springer　Nature　Singapore　Pte　Ltd.