- SuperTarget^76 is a comprehensive drug-target database obtained by combining
various drug-related information, such as adverse drug effects, drug metabolism,
pathways, and Gene Ontology terms of the target proteins. - BRaunschweig ENzyme DAtabase (BRENDA)^77 is one of the most comprehen-
sive enzyme repository, which contains molecular and biochemical information
on enzymes.
3.1.8 Drug Interaction Networks
There are many ways to construct a drug-drug interaction network. As mentioned in
the previous subsection, a drug-drug interaction network can easily be obtained by
projecting the drug-target bipartite network onto the drug-drug interaction network
by connecting two drugs if they share at least one target protein.^78 However, many
previous approaches exploit many other data types, such as the chemical structure
data and information on drug side effects to construct informative drug-drug
interaction networks. These approaches can be divided into three different
classes:^79
- Chemical structure similarity between drugs—information on chemical struc-
ture of drugs can be obtained from DrugBank database^80 or KEGG LIGAND
database.^81 Based on the molecular structure of drugs, various measures can be
used for computing drug similarity, such as Jaccard Index (JI), the Cosine
Similarity (CS), or the Dice Coefficient (DC). - Side-effect similarity between drugs—clinical side effects provide a human
phenotype profile for a drug. Information on drugs’side effects can be obtained
from the Side Effect Resource (SIDER) database (Table 1 ).^82 Each drug can be
represented asn-dimensional binary vector representing side-effect profile with
elements 0 or 1 that encode for the presence or absence of the side-effect key
words. By having these vectors, it is possible to define a pairwise side-effect
similarity between two drugs using the Jaccard Index.^83 - Gene expression similarity as a response to a drug’s action—gene expression
profiles under the influence of drugs can be retrieved from the Connectivity Map
(CMAP) project database (Table 1 ).^84
(^76) Gunther et al. ( 2008 ).
(^77) Schomburg et al. ( 2013 ).
(^78) Yildirim et al. ( 2007 ).
(^79) Ding et al. ( 2013 ).
(^80) Wishart et al. ( 2008 ).
(^81) Kanehisa et al. ( 2006 ).
(^82) Kuhn et al. ( 2010 ).
(^83) Zhang et al. ( 2013 ).
(^84) Lamb et al. ( 2006 ).
152 V. Gligorijevic ́and N. Pržulj