Prediction Of Drug-Target Modules Based On Multilayer Network Models

A drug target is a specific molecule in a cell that binds to the drug and makes the drug have some effects directly,and more than 98% of drug targets belong to proteins.The role of drugs in combination with target proteins is embodied in complex biological processes.At present,most researches on drug target methods are based on global single-layer PPI network.However,in most cases,one drug can treat many diseases and each disease only affects gene expression in related tissues.The layer PPI network does not take into account the tissue specificity of the disease,leading to bias in the study of drug targets.To date,most people have devoted themselves to studying the relationship between drugs and individual targets.Normally,genes do not act alone,but interact with other genes in functional modules to achieve specific functions.Therefore,drugs can be used to treat diseases by acting on functional modules,and these functional modules are called drug-target modules.We propose a new framework for predicting drug-target modules.First,we construct a set of genes where a drug affected in different diseases actually.Then,we screen out the gene-generating subgraphs of public genes in tissue-specific PPI networks related to diseases,and build a multi-layer network with no borders between layers.Afterwards,a multi-layer network module mining algorithm is used to mine all possible candidate drug-target modules.And then,gene set that the drug truly affected are mapped into candidate drug-target modules and the drug's possible target modules are initially screened.After that,GO terms of modules are enriched,and overlapped with enrichment results of the drug targets and disease-causing genes in order to screen possible drug-target modules.Finally,KEGG pathways of modules are enriched,and overlapped with enrichment results of the drug targets to screen final drug-target modules.Through the analysis of the significances and functions of the selected drug-target modules,this paper validates the feasibility and rationality of this framework.Based on the above framework,two potential target modules are predicted based on three diseases of TSA treated,including leukemia,breast cancer and prostate cancer.This framework breaks the previous prediction approach centered on individual drug targets and proposes finding drug-target modules in a multi-layer network,focusing more on the functional mechanism of the drug,making up for the deficiencies brought about by the single-layer network as a background and identifying the functions of drugs at the level of biological tissues associated with complex diseases.The research of the drug target module prediction method based on the multi-layer network model is helpful for us to understand the functions of drugs from a brand-new and more comprehensive perspective and provide a new direction for the treatment of diseases.