About this Web Site
This web-site was developed as a prototype to support building drug-drug similarity networks and connecting drugs to the human interactome through known drug targets. A manuscript describing the project with case studies and global analyses has been submitted for publication:
Drugs2Networks: Linking Approved Drugs Based on their Shared Properties. Submitted
About the Drug-Drug Similarity Networks
To connect drugs or properties of drugs, Drugs2Networks (D2N) uses an underlying knowledge base made of six drug-drug interaction networks. These networks are created by connecting drugs if pairs of drugs share significant AERS adverse events, or SIDER side effects; Connections are also made if pairs of drugs share similar gene expression signature in CMAP. To reduce the dimensionality of the drug-drug similarity from CMAP we converted sets of differentially expressed genes for each drug to enriched pathways from the KEGG pathway database, or by using our own ChIP-X Enrichment Analysis database (ChEA); We also created a drug-drug similarity network by connecting two drugs if they significantly share ATC codes; The next drug-drug network is based on structural similarity, where similarity is determined based on the drug's 166-key MACCS fingerprint created from the drugs' SMILES strings using ChemmineR. Finally, the last drug-drug network is based on shared targets listed in DrugBank. The AERS, SIDER, KEGG, ChEA, ATC and MACCS drug-drug networks are made based on the shared "properties" of the drugs. These properties can also be connected based on the drugs they share. Drugs2Networks (D2N) contain six property-property networks and users can choose a property to build property-property similarity subnetworks.
The six drug-drug similarity networks and the six property-property similarity networks are created from Gene Matrix Transpose (GMT) files that store six different drug-set libraries: AERS, SIDER, KEGG, ChEA, ATC and MACCS. After creating these six GMT files we used our algorithm Sets2Networks [PubMed ID: 22824380] to convert the GMT files into networks. These six GMT drug-set libraries are available for download from here:
- AERS GMT file
- SIDER GMT file
- CMAP-ChEA GMT file
- CMAP-KEGG GMT file
- ATC codes GMT file
- MACCS GMT file
About the Drug-Target, Disease Gene, Virus-Protein-Human-Protein, and Protein-Protein Networks
To create subnetworks that connect drug targets, viral proteins, and disease genes through known protein-protein interactions we constructed drug-target, disease-gene and virus-protein/human-protein bipartite networks. The drug-target interactions are originated from DrugBank; the disease gene lists are derived from OMIM; and the virus-host interactions are derived from VirusMINT. From these sources we obtained bipartite networks where one side of the graph is represented by drugs, diseases, or viral proteins, and the other human genes. The human genes from each of these categories are then connected through known protein-protein interactions collected from 18 publicly available databases. We use a shortest path algorithm we previously implemented for the program Genes2Networks [PubMed ID: 17916244].
Contact
Drugs2Networks was designed and implemented by the Ma'ayan laboratory at the Mount Sinai School of Medicine in New York. Questions and comments about this work should be directed to avisomething.morestuffmaayan@nospammssm.edu