for many different species by RNA microarray technology, which enabled massive
parallel measurement of expressions of thousands of genes simultaneously.^53 To
construct a gene coexpression network, we first need to construct a correlation
matrix of gene coexpressions. This square matrix has the size equal to the number
of genes in an organism, and its entries represent Pearson correlation coefficients
(PCCs) of expression profiles between all pairs of genes. To construct a gene
coexpression network from this matrix, we need to carefully choose an appropriate
threshold: we link a pair of genes if their PCC exceeds that threshold.^54 Evidently,
network topology will greatly depend on the choice of the threshold. Therefore, it is
essential to choose an appropriate threshold. There is no agreed-upon methodology
for determining an appropriate threshold. Some of the methods for obtaining the
right threshold are based on the known biological information,^55 statistical com-
parison with randomized expression data,^56 or random matrix theory (RMT).^57
The most widely used data repositories (see Table 1 ) containing gene expression
profiles are Gene Expression Omnibus (GEO),^58 ArrayExpress,^59 and Stanford
Microarray Database (SMD).^60
3.1.4 Genetic Interaction (GI) Networks
A genetic interaction is a binary interaction between two genes such that the double
deletion (mutation) leads to a change in the organism’sphenotype(observable
characteristics of an organism) and single deletion of either one of the two genes
do not.^61 The most commonly used measure of the phenotype is the number of
cells in the culture. Two types of genetic interactions can be defined based on the
mutual effect of the two genes on the phenotype. Two genes are said to exhibit a
positive (negative) genetic interaction if their mutual deletion leads to a signifi-
cantly worse (better) phenotype than expected from single gene deletions of two
genes under the multiplicative rule.^62 The comparison of the phenotypes can be
performed by evaluating the difference in the number of cells in the culture. The
positive genetic interaction is also known as asynthetic sickorlethal, while the
negative genetic interaction is also known assuppressive.^63 In a genetic interaction
(^53) Luo et al. ( 2007 ).
(^54) Prieto et al. ( 2008 ).
(^55) Ziv et al. ( 2003 ).
(^56) De Smet et al. ( 2002 ).
(^57) Luo et al. ( 2007 ).
(^58) Barrett et al. ( 2007 ).
(^59) Parkinson et al. ( 2005 ).
(^60) Hubble et al. ( 2009 ).
(^61) Costanzo et al. ( 2010 ), Mani et al. ( 2008 ), and Vidal et al. ( 2011 ).
(^62) Mani et al. ( 2008 ).
(^63) Vidal et al. ( 2011 ).
Computational Methods for Integration of Biological Data 149