system-level analysis. This is because the use of network formalism enabled us to
apply a wide variety of mathematical tools to these data that have yielded new
biological insight. Many diseases such as cancer, diabetes, and cardiovascular
diseases are not caused by single genes, but rather they are a result of a very complex
network machinery of interacting molecules and pathways. In recent years, much
attention has been paid to developing computational and mathematical tools for
analysis and modeling of a single individual molecular network type (i.e.,omics
layer). For a recent review of network-based methodologies for molecular network
analyses, see Pržulj ( 2011 ), Sarajlic ́ and Pržulj ( 2014 ) and Aittokallio and
Schwikowski ( 2006 ).^7 However, individual analyses of each particularomicslayer
in isolation can reveal only partial biological knowledge and cannot decipher the
complete heterogeneous interaction landscape of a cell as these networks are often
incomplete (e.g., many interactions are missing) and they represent only a partial
view of a cell’s molecular structure. Despite these shortcomings, previous analyses
of each separate network layer have provided valuable biological information.
Integration ofomicslayers increases the coverage of the available data and reveals
additional, hidden biological patterns, the analysis of which could contribute to
addressing many relevant biological questions. Findings obtained from these inte-
grative analyses could further be translated into medical practice.
In addition to experimentally extracted interaction data, there is a large amount
of other diverse biological data available in many repositories. Such data often
represent a valuable complement to the interaction data in computational extraction
of new biological knowledge. These data include genetic data obtained from
genome-wide association studies (GWAS),^8 sequence data, functional annotation
data, and ontology data (such as Gene Ontology^9 and Disease Ontology^10 ). These
data can be interpreted as additional features of biological entities (genes, proteins,
diseases, drugs, etc.), and they can be easily incorporated in an integration frame-
work, along with theomicsdata, to increase the reliability of newly discovered
knowledge.^11 Some examples include integration of the interactome network with
sequence, functional annotation, and expression data to predict new protein-protein
interactions in yeast; drug repositioning by integrating information on drug chem-
ical structure, drug target and side effects;^12 predicting protein function by inte-
grating molecular networks with other data, such as molecular sequence, protein
domains, and gene expression profiles;^13 predicting new candidate genes for breast
cancer by integrating gene coexpression network associated with breast cancer
genes, along with genetic interaction network and PPI network.^14
(^7) Pržulj ( 2011 ), Sarajlic ́and Pržulj ( 2014 ), Aittokallio and Schwikowski ( 2006 ).
(^8) Mccarthy et al. ( 2008 ), Sladek et al. ( 2007 ), and The Wellcome Trust Case Control
Consortium ( 2007 ).
(^9) Ashburner et al. ( 2000 ).
(^10) Schriml et al. ( 2012 ).
(^11) Lu et al. ( 2005 ).
(^12) Wang et al. ( 2013 ).
(^13) Ma et al. ( 2013 ).
(^14) Zhang et al. ( 2011 ).
Computational Methods for Integration of Biological Data 139