been used (Table 1 ): General Repository for Interaction Datasets (BioGRID),^48
Search Tool for the Retrieval of Interacting Genes Database (STRING),^49 and
Human Protein Reference Database (HPRD).^50 STRING database contains many
interactions that are computationally predicted, and these interactions are often
excluded in most studies. On the other hand, BioGRID contains only curated
physical and genetic interactions. When constructing a PPI network using data
sets from BioGRID, it is a common practice to consider only highly significant
interactions (e.g., those that are reported in two or more different studies).
3.1.2 Metabolic Networks
A metabolic network includes all possible metabolic reactions within an organ-
ism.^51 The easiest way to represent a metabolic network is by using a bipartite graph
representation: nodes in one partition represent metabolites, while nodes in other
partition represent possible reactions. An edge between a metabolite and a reaction
exists if the metabolite participates in the reaction. The main actor of each reaction
is an enzyme, the primary role of which is to catalyze the reaction. Most enzymes
are large proteins. Another way to represent a metabolic network is by replacing
reactions with enzymes that catalyze them. This representation is more universal
since it provides a possibility for comparison of metabolic networks between
different species. For example, if genes that produce particular enzymes do not
exist in a species’genome, the corresponding reaction does not occur within that
species. A more general case is to replace each enzyme with a gene that codes for
that enzyme. In this way we obtain a metabolite-gene bipartite network. It is a
common practice in many data integration studies to project a bipartite network
representation of a metabolic network onto a simple gene-gene network represen-
tation. Such a representation allows a much easier integration of metabolic network
with other molecular networks (also represented in the gene-gene form). The usual
way to do this projection is by removing all metabolites and connecting two genes if
they are connected with the same metabolite in the bipartite network representation.
The most widely used repository of metabolic networks is Kyoto Encyclopedia of
Genes and Genomes (KEGG).^52
3.1.3 Gene Coexpression (Co-Ex) Networks
Gene expression is a process by which gene information is used to synthesize gene
products (proteins). Vast amounts of gene coexpression data have been generated
(^48) Chatr-Aryamontri et al. ( 2013 ).
(^49) Franceschini et al. ( 2013 ).
(^50) Keshava Prasad et al. ( 2009 ).
(^51) Schuster et al. ( 2000 ) and Vidal et al. ( 2011 ).
(^52) Kanehisa et al. ( 2012 ) and Zhou ( 2013 ).
148 V. Gligorijevic ́and N. Pržulj