Systems Biology (Methods in Molecular Biology)

DEGs network using the information available on MirTarbase [92],

a database for experimentally validated miRNA target genes,

miRanda [93] and TargetScan [94], two databases for predicted

miRNA target genes. For each of the DETF-DEG pairs, we calcu-

lated the Pearson correlation coefficient (PCC) of expression values

and selected the significant pairs with an absolute PCC value more

than 0.4 (as TF can either trans-activate or trans-suppress their

target genes). The main regulatory effect of miRNAs is induced

through binding and degradation of their target mRNA. There-

fore, we considered only negative correlation between miRNA and

the target mRNA expression [95–97] and set the PCC cutoff to

<
0.4. This filtering significantly reduced number of pairs for the

construction of regulatory network. Finally, from the selected

DEM-DEG and DETF-DEG pairs, we constructed and visualized

a composite highly differentially regulated co-expression network

for each clinical stage of prostate cancer progression using Cytos-

cape [13] as shown in Fig. 10. The number of nodes in the

MAOB

EGR1

SNAI2

MYC

ITGA3

ZFP36

GDF15

miR-22*

miR-575

ZEB1

ACTA2

DES

FBXO32

TEAD1

miR-200c

miR-627

miR-548c-3p

miR-629

miR-25 GLI3

BCL2

ATP1B1

EDNRA

TP63

PGR

miR-135a*

HOXD10

miR-32

IGFBP3

miR-375

miR-19a

miR-19b miR-20a*

ELF1 miR-663

miR-494 miR-665

miR-218 miR-145

BIRC5

miR-512-3p

BHLHE40

CYR61 SERPINA3

IL6ST

miR-671-5p

EPAS1 MEIS1 PROS1

IGFBP4 PTGDS STAT3 SOCS3 PTTG1

HLA-B ZEB1

GBP1

CEACAM1

miR-30d SNAI2

miR-371-5p MMP2

JUNB

ITGA5

PTGS2 FOS

THBS1 let-7a*

IRF1

miR-1237

KIT

miR-1228

ID2

BTG2

NFIA

FBXO32

TGFBR2

ZFP36

ETS2

ATF3

FOSB

miR-769-5p

LDLR

EGR1

SELE

NR4A2

miR-125a-3p

JUN

TSC22D1

miR-377

NR4A1

miR-205

miR-33b

miR-143

miR-29b

miR-27b

NFIB

miR-29a

ITGA6

CCL2

let-7e

TOP2A

miR-28-5p

AR

BRIP1

BRCA1

EZH2

miR-199a-5p

E2F5

CDK1

miR-1

SMAD9

NDRG2

CDC6

MYBL2 FOXM1

KLF4

SMAD4

SEPP1

TP63

RGS2

GADD45B

CTGF

GDF15

CEBPD

miR-30c miR-29c

KRT14

(b)

(a)

-1.50 1.50

log2 fold change

Fig. 10Integrative network constructed using significant pairs of differentially expressed gene, transcription
factors, and miRNAs based on the co-expression correlation of connected nodes for (a) primary; and (b)
metastatic prostate cancer phenotype. MiRNA, genes, and TFs are represented as oval, hexagonal, and
diamond shape nodes respectively. Nodes are colored based on their log2 fold change expression values
(green: down-regulated;red: up-regulated). Edge color indicates type of regulation (red: activation;blue:
suppression) and edge width is proportional to the absolute correlation coefficient for the expression values of
connected pairs

266 Faiz M. Khan et al.