Computational Systems Biology Methods and Protocols.7z

gxa), an integrative expression database providing differential and

baseline expression information [6]. On the other hand, the gene

expression data can be produced de novo through high-

throughput omics methods, such as microarray technology,

RNA-seq assay, and RT-PCR experiment. Please keep in mind

that sample size of expression data is an important consideration

for GRN reconstruction since various inference methods have dif-

ferent size requirement. In the following sections, we will show

how these network methods work and how they can be applied in

exploring disease mechanism.

2 Inference Methods of GRN Reconstruction

Recently, enormous network inference methods have been devel-

oped in computational biology field. These methods can be

grouped into two categories, one for static network and another

for dynamic network. In here, the static network means a GRN is

constructed without limitation of spatial and temporal conditions,

while a dynamic network describes a GRN under a spatiotemporal

condition.

2.1 Methods
for Static Networks

2.1.1 Information-
Theoretic Models

A particular advantage of information-theoretic model is its mini-

mal mathematical assumptions of network reconstruction

[7, 8]. However, an information-theoretic model can only provide

an undirected network commonly. For this model, a correlated

measurement is calculated to capture correlation between genes

for network inference. The most popular correlated measurements

are the Pearson’s correlation coefficient (PCC) and the Spearman’s

rank correlation coefficient (SCC) [9, 10]. While the former can

detect linear correlation, the latter is suitable to nonlinear correla-

tion inspection. The PCC and SCC measurements between genes X

and Y can be expressed as follows.

PCC XðÞ¼;Y

covðÞX;Y

σXσY

¼

Pn

i¼ 1

ðÞXiX YiY



ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

Pn

i¼ 1

ðÞXiX

2

s ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

Pn

i¼ 1

YiY

 2

s

SCC XðÞ;Y ¼

cov Rx;Ry



σRxσRy

ð 1 Þ

wherenis the number of experiment andidenotes theith expres-

sion value of a gene, cov(X,Y) is covariance between genes X and Y,

cov(Rx,Ry) is covariance between rank variablesXandY,σXandσY

138 Guangyong Zheng and Tao Huang