Drug Repositioning Based On Phenotypes And Pathways

Traditionally, drug discovery process includes three stages: discovery, preclinical and clinical development stage. The whole process is time-consuming and expensive, but often inefficient. Drug repositioning is proposed in this case, the goal is to identify and use known drugs that can treat diseases other than those for which they were originally designed. Conventional drugs repositioning methods are mainly based on the relationship of drugs genes diseases, and discarded a number of important information. In order to solve this problem, we integrate a variety of data, from many angles to infer drug function, we propose two new methods to addressing the drug repositioning problem.The first approach focuses on phenotypic data, presents a forecast based on side effects of drugs and clinical symptoms of the disease to predict a drug- disease correlation. Firstly, using inverse document frequency, a drug-drug similarity network and a disease-disease similarity network is constructed based on side effects of drugs, symptoms of diseases. Using C luster One to cluster the two networks respectively, we get two sets of modules, drug- module sets and disease- module sets. According to the known associations between drugs and diseases in CTD database, we correlate drug modules and disease modules and score the correlations between each pair of drug-disease module. Finally, we rank the drug-disease module pairs according to their correlations. We select the top 3 drug-disease module pairs to analyze and predict the unknown association between the drug and the disease. We evaluate our prediction with drug-disease associations that are reported in published literature and Comparative Toxicogenomics Database(C TD; http://ctd.mdibl.org) benchmark, and we also make K yoto Encyclopedia of Genes and Genomes(K EGG) enrichment analysis on their related genes.The second method proposeda method to build a disease- disease similar network(Human Pathways-Disease Network, HPDN)based on pathway from a functional pathway perspective, and the relocation HPDN used in pharmaceutical research. First, based on the path to obtain the relevant sets of genetic diseases. Then, based o n the gene sets distance measure calculated the correlation between each pair of diseases, building disease- disease similar networks. Finally, compare the random network, filtering disease- disease similar network, and ultimately build a network(HPDN) for authentication. The HPDN compared with a network of the two diseases based on genes and based on symptoms and found to have a high similarity between them, and HPDN can predict new disease-disease correlation, and can be effectively used in drug relocation.In summary, the paper not only provides an effective strategy to predict repositoning drug based drug-disease network, but also provides a new idea for related disease research, providing an important basis for the study of the drug repositioning.