Network Biology

Vol. 4, No. 1, 1 March 2014

International Academy of Ecology and Environmental Sciences

Network Biology
ISSN 2220-8879
Volume 4, Number 1, 1 March 2014

Editor-in-Chief
WenJun Zhang
Sun Yat-sen University, China
International Academy of Ecology and Environmental Sciences, Hong Kong
E-mail: zhwj@mail.sysu.edu.cn, wjzhang@iaees.org

Editorial Board
Ronaldo Angelini (The Federal University of Rio Grande do Norte, Brazil)
Sudin Bhattacharya (The Hamner Institutes for Health Sciences, USA)
Andre Bianconi (Sao Paulo State University (Unesp), Brazil)
Danail Bonchev (Virginia Commonwealth University, USA)
Graeme Boswell (University of Glamorgan, UK)
Jake Chen (Indiana University-Purdue University Indianapolis, USA)
Ming Chen (Zhejiang University, China)
Daniela Cianelli (University of Naples Parthenope, Italy)
Kurt Fellenberg (Technische Universitaet Muenchen, Germany)
Alessandro Ferrarini (University of Parma, Italy)
Vadim Fraifeld (Ben-Gurion University of the Negev, Israel)
Alberto de la Fuente (CRS4, Italy)
Pietro Hiram Guzzi (University Magna Graecia of Catanzaro, Italy)
Yongqun He (University of Michigan, USA)
Shruti Jain (Jaypee University of Information Technology, India)
Sarath Chandra Janga (University of Illinois at Urbana-Champaign, USA)
Istvan Karsai (East Tennessee State University, USA)
Caner Kazanci (University of Georgia, USA)
Vladimir Krivtsov (Heriot-Watt University, UK)
Miguel ángel Medina (Universidad de Málaga, Spain)
Lev V. Nedorezov (Russian Academy of Sciences, Russia)
Alexandre Ferreira Ramos (University of Sao Paulo, Brazil)
Santanu Ray (Visva Bharati University, India)
Dimitrios Roukos(Ioannina University School of Medicine, Greece)
Ronald Taylor (Pacific Northwest National Laboratory,U.S. Dept of Energy, USA)
Ezio Venturino (Universita’ di Torino, Italy)
Jason Jianhua Xuan (Virginia Polytechnic Institute and State University, USA)
Ming Zhan (National Institute on Aging, NIH, USA)
TianShou Zhou (Sun Yat-Sen University, China)

Editorial Office: networkbiology@iaees.org

Publisher: International Academy of Ecology and Environmental Sciences

Address: Flat C, 23/F, Lucky Plaza, 315-321 Lockhart Road, Wanchai, Hong Kong
Tel: 00852-6555 7188
Fax: 00852-3177 9906
Website: http://www.iaees.org/
E-mail: office@iaees.org
Network Biology, 2014, 4(1): 1-20

Article

Meta-analysis of cancer transcriptomes: A new approach to

uncover molecular pathological events in different cancer tissues
Sundus Iqbal1, Hira Ejaz1, Muhammad Sulaman Nawaz1, Juan J Loor2, Aisha Naeem1
1
COMSATS Institute of Information Technology, ChakShahzad, Park Road, Islamabad, 44000, Pakistan
2
Mammalian NutriPhysioGenomics, Department of Animal Sciences and Division of Nutritional Sciences, University of
Illinois, Urbana, IL 61801, USA
E-mail: aisha_naeem@comsats.edu.pk

Received 23 October 2013; Accepted 28 November 2013; Published online 1 March 2014

Abstract
To explore secrets of metastatic cancers, individual expression of true sets of respective genes must spread
across the tissue. In this study, meta-analysis for transcriptional profiles of oncogenes was carried out to hunt
critical genes or networks helping in metastasizing cancers. For this, transcriptomic analysis of different
cancerous tissues causing leukemia, lung, liver, spleen, colorectal, colon, breast, bladder, and kidney cancers
was performed by extracting microarray expression data from online resource; Gene Expression Omnibus. A
newly developed bioinformatics technique; Dynamic Impact Approach (DIA) was applied for enrichment
analysis of transcriptional profiles using Database for Annotation Visualization and Integrated Discovery
(DAVID). Furthermore, oPOSSUM (v. 2.0) and Cytoscape (v. 2.8.2) were used for in-depth analysis of
transcription factors and regulatory gene networks respectively. DAVID analysis uncovered the most
significantly enriched pathways in molecular functions that were ‘Ubiquitin thiolesterase activity’ up
regulated in blood, breast, bladder, colorectal, lung, spleen, prostrate cancer. ‘Transforming growth factor
beta receptor activity’ was inhibited in all cancers except leukemia, colon and liver cancer. oPOSSUM
further revealed highly over-represented Transcription Factors (TFs); Broad-complex_3, Broad-complex_4,
and Foxd3 except for leukemia and bladder cancer. From these findings, it is possible to target genes and
networks, play a crucial role in the development of cancer. In the future, these transcription factors can serve
as potential candidates for the therapeutic drug targets which can impede the deadly spread.

Keywords metastatic cancer; transcriptional analysis; oPOSSUM; transcription factor; DAVID.

Network Biology
ISSN 2220­8879
URL: http://www.iaees.org/publications/journals/nb/online­version.asp
RSS: http://www.iaees.org/publications/journals/nb/rss.xml
E­mail: networkbiology@iaees.org
Editor­in­Chief: WenJun Zhang
Publisher: International Academy of Ecology and Environmental Sciences

1 Introduction
Cancer is not a single disease, it is a phrase used to identify immense number of similar diseases resulting
from the interplay of gene(s) and environmental factors (Emilsson et al., 2008). Although, cancer can affect

IAEES www.iaees.org
2 Network Biology, 2014, 4(1): 1-20

every organ or tissue in the human body, the basic pathology remains the same. As abnormal proliferation
and failure to cell death in cancer occurs due to the accumulation of mutations in oncogenes or tumor
suppressor genes.
The great majority of cancer deaths occur due to the ability of the cancer to metastasize to other organs.
Over the last century, biological research has generated a wealth of knowledge about the cellular structural
and functional attributes of cancer that lead to initiation and metastasis. Cancerous cells acquire metastatic
abilities due to alterations that surpass their physiological barriers as they separate from their original
developmental fate and environment (Nguyen and Massague, 2007). Thus, tumor cells proliferate and
penetrate into new tissues, which eventually result in organ dysfunction and death. Understanding the
metastatic process is far more challenging than early phases of cancer due to its complex cell-to-cell
interaction and microenvironment that promote the process (Bardelli et al., 2003). Application of
bioinformatics can help to uncover the molecular processes involved in metastatic cancer in a more systems
oriented fashion.
Gene expression profiling and other high throughput technologies have helped in understanding and
performing systematic analyses of many complex diseases (Schadt et al., 2005). Microarrays put forth a
platform to identify biomarkers along with mechanisms of toxicity and pathogenicity (Waring et al., 2002)
and disease subtypes (Mootha et al., 2003; Schadt et al., 2003; van 't Veer et al., 2002). Bioinformatics tools
further allow to re-construct gene networks by the integration of genetic data with gene expression to
decipher the cellular dynamics during tumorigenesis (Zhu et al., 2004).
Rhodes et al. (2004) have attempted to recognize cancer type-independent gene expression signatures and
their corresponding transcriptional factors by comparative meta-profiling of microarrays of a wide range of
cancers. They have characterized universally regulated transcriptional profiles in both well-differentiated and
undifferentiated cancers. In another study, Wang et al. (2009) have identified gene-to-gene co-expression
networks in liver among multiple species by performing semi-parametric meta-analysis strategy to study
relationships among common human diseases. Using meta-profiling, researchers have revealed that down
regulated genes in Duchenne Muscular Dystrophy (DMD) are involved in a single shared pathway that is
responsible for creating disruption in activity of muscle related transcription factors (TF), which contribute to
the severity of DMD and these TF can be potential candidates for drug targets (Kotelnikovaet al., 2012).
In this study, we have identified tissue specific and conserved molecular pathways in different metastatic
cancerous tissues e.g., lung, prostate, colorectal, kidney, breast, spleen, liver, colon, bone and bladder based
on freely available gene expression data of these cancers from the GEO data base using the newly-developed
‘Dynamic Impact Approach (DIA) (Bionaz et al., 2012). Furthermore, for in-depth analysis of the identified
molecular mechanisms, their association with differentially expressed TFs was evaluated as a means to
potentially uncover potential drug targets (Blancafortet al., 2004; Dunker andUversky, 2010; Smith andBirrer,
1996). We speculate that simultaneous analysis of pathological events of cancers in different tissues can
reveal conserved networks across different tissues that can highlight the main biological processes involved in
pathogenesis and metastatic abilities of cancer.

2 Methods
Gene Expression Omnibus database (GEO) at National Centre for Biotechnology Information (NCBI) stores
high throughput functional genomics datasets (Edgar and Barrett, 2006) and was created by adopting both
microarray and sequence technologies. For this study, the data for different cancers were extracted from GEO
on the basis of population, gender, age, pathological grade and primary site (Online material, file S1). All
selected microarray experiments were performed on human genome U133 plus 2.0. The list of differentially

IAEES www.iaees.org
 Network Biology, 2014, 4(1): 1-20 3

expressed genes (DEGs) with the cut off P-value was <0.001 and FDR 0.05 was used for all the down stream
analysis. All the supplementary information is available at http://www.ncbi.nih.gov/geo of colon
(GSM38055), lung(GSM38058), spleen (GSM38056), liver (GSM46848), breast (GSM38109), papillary
renal cell carcinoma (GSM46847) and prostate cancer. The clinical parameters of the samples are reported in
Table S1, file S1 (Supplementary material). The data for bladder leukemia and bladder cancer was extracted
from Savli et al. (2012) and Zhu et al. (2011).
The scheme of the study is presented in Fig. 1. The data analysis was divided into three parts, which
include the following: (1) Enrichment Analysis by DIA (2) Transcription Factor Analysis performed by using
the web server oPOSSUM (v.2.0) and (3) Gene Regulatory Networks generated by Cytoscape (v.2.8.2).
Enrichment analysis was performed by using the DIA with the help of a freely available webserver, DAVID
(v.6.7) (The Database for Annotation, Visualization and Integrated Discovery (DAVID version 6.7; http://
david.abcc.ncifcrf.gov/) to identify affected pathways, biological processes, and molecular functions within
the DEGs. The detailed method for the analysis has been published previously (Bionaz et al., 2012). Briefly,
the entire microarray data sets with associated statistical P-values were imported into DAVID. The ‘Impact’
refers to the absolute perturbation in a biological process (i.e., overall dynamics within a term/pathway). It
gives the magnitude of change that occur (due to treatment) within a pathway/function in either direction
weighted by the percent DEGs that hit the term/pathway. The ‘Flux’ (activation or inhibition) refers to the
overall direction in which a term/pathway is impacted after treatment (Bionaz et al., 2012). The length of the
bars depicts the degree of the impact, and the intensity of the color (e.g., from dark green denoting highly
down regulated to dark red denoting highly upregulated) was used to highlight the degree of activation or
inhibition of pathways and terms. As DIA is not a statistical approach; thus, it is not possible to apply cut-offs
in order to determine if terms have a biological significance. Furthermore, the results (i.e., the impact) can be
considered an indication of the effect of changes in the transcriptome on the biological terms. In this sense,
the impact between two terms can be compared (i.e., one term being more impacted than another) in a
relative fashion. Because the impact is an ‘absolute’ value, the same term also can be compared between two
different experiments. As an example, biological terms that have an impact over 100 were considered as
substantially affected by the treatment (i.e., a pathway with an impact of 100 needs to have on average 25%
of protein-genes being significantly affected with a minimum of 2-fold change and 10−4 P-value) (Bionaz et
al., 2012).
For the analysis of sequences and key regulatory networks that regulate transcriptional responses under
specific environment, oPOSSUM version 2.0 Human Site Analysis (link: http://www.cisreg.ca/cgi-
bin/oPOSSUM/opossum/) (Huang et al., 2006) and Cytoscape (Lopes et al., 2010) were used. oPOSSUM has
been reported to identify the TFs that can mediate changes in gene expression through the detection of over-
represented TFBSs in a set of differentially expressed genes in comparison with the pre-compiled background
gene sets. The identification of over-represented transcription factor binding sites (TFBS) from sets of co-
expressed genes provides insights into the mechanisms of regulation for diverse biological contexts.
oPOSSUM calculates two statistical measures for binding site over-representation, one at the gene level
(Fisher exact test) and the other based on the ratio of TFBSs to nucleotides (Z-score). oPOSSUM accepts
input in the form of gene identifiers e.g. Entrez ID. Against the background set, oPOSSUM compares number
of hits for each selected TFBS in the target gene set. The database applies two statistical significance
measures, which are Z-score and Fisher-score. Target gene hits and their respective TFs are retrieved through
oPOSSUM. Cytoscape version 2.8.2 was used to create networks of the over expressed TF and their target
gene list for the network analysis.

IAEES www.iaees.org
4 Network Biology, 2014, 4(1): 1-20

Fig. 1 Scheme of the study.

3 Results
Meta-analysis of transcriptional profiles of different cancers was performed to examine the expression of
genes in response to cancer. A total of differential upregulated and downregulated genes with the cutoff as
defined above fromselected microarray experiments are presented in Table 1. The aggregated number of
DEGs identified in blood, lung, prostate, colorectal, papillary renal, breast, spleen, liver, colon and bladder
cancers was160, 1015, 314, 610, 1823, 307, 309, 278, 220, 2018, respectively. For computational analysis
with the DIA approach, we used all the 28 categories within DAVID but only Molecular Function
(GOTERM_ MF_FAT) are reported here (Online material, file S2). Details of these categories were already
reported elsewhere (Bionaz et al., 2012).

Table 1 Number of differentially expressed genes in different cancer tissues with cut offp-value<0.001 and FDR 0.05.

DEGs Lung Prostate Colorectal Renal Breast Spleen Liver Colon Leukemia Bladder
Up 200 200 312 201 201 201 201 201 71 1026
Down 815 114 298 1622 106 108 77 19 89 992
Total 1015 314 610 1823 307 309 278 220 160 2018

With the cut off value stated above, our analysis clearly showed that cancer in all the different tissues
sharesseveral molecular events. The most enriched molecular functionsin all the selected cancer profiles are
reported in Fig. 2. A global pattern of transcriptional responses in all the cancers, as determined from analysis
of cancers in different tissues is shown in file S1 in the form of heat map. The top-most GO terms uncovered
by DIAwere ‘Extracellular matrix structural constituent’, ‘Lipase inhibitor activity’, ‘Collagen binding’,
‘Integrin binding’, Metalloendopeptidase inhibitor activity’, ‘Metalloenzyme regulator, ‘Metalloenzyme

IAEES www.iaees.org
 Network Biology, 2014, 4(1): 1-20 5

inhibitor’‘Fibroblast growth factor 2 binding’, ‘Fibrinogen binding’, ‘Oxidoreductase activity, NAD or

NADP’, ‘Alcohol dehydrogenase activity, zinc dependent’ and ‘Fibronectin binding’ (Fig. 2).

Molecular Functions Lung Prostate Colorectal Renal Breast Spleen Liver Colon Leukeima

extracellular matrix structural constituent
lipase inhibitor activity
cyclin‐dependent protein kinase inhibitor activity
extracellular matrix structural constituent conferring tensile strength
fibronectin binding
collagen binding
integrin binding
metalloendopeptidase inhibitor activity
metalloenzyme regulator activity
metalloenzyme inhibitor activity
glycoprotein binding
opsonin binding
cytoskeletal adaptor activity
oxidoreductase activity, acting on the CH‐NH2 group of donors
heparin binding
low‐density lipoprotein receptor activity
glycosaminoglycan binding
WW domain binding
sulfuric ester hydrolase activity
transforming growth factor beta binding
pattern binding
calcium‐dependent phospholipid binding
snRNA binding
oxidoreductase activity, acting on the CH‐NH2 group of donors
oxygen transporter activity
transaminase activity
serine‐type endopeptidase inhibitor activity
peptidase inhibitor activity
peroxidase activity
oxidoreductase activity, acting on peroxide as acceptor
antigen binding
glutathione transferase activity
small conjugating protein binding
lipoprotein binding
lipopolysaccharide binding
oxygen binding
adenyl nucleotide binding
transcriptioncorepressor activity
endonuclease activity
insulin‐like growth factor receptor binding
amino acid binding
acid‐amino acid ligase activity

Fig. 2 The most impacted gene ontology terms (GO)as depicted by Dynamic Impact Approach (DIA) in different cancer
tissues.On the extreme left are reported the most impacted biological terms. On the right of the column with the biological term
is the column of the each tissue typeconsidered in this study. The column presents two sub-columns. In the left sub-column it is
reported the horizontal blue bar that denotes the overall impact of the differentially expressed genes on the biological term.
Larger the horizontal bar larger the impact. In the right sub-column it is reported a colored square that denotes the direction of
the impact (green=inhibited; red=activated). Darker the color larger the activation (if red) or inhibition (if green) of the
biological term.

The most impacted GO term identified in all cancers was ‘Extracellular matrix structural constituent’.
‘Extracellular matrix structural constituent’ is strongly induced in lung, breast spleen liver and colon cancer
while inhibited in others. ‘Collagen binding’ that is involved in induction of cellular proliferation in many
cancers is namely up regulatedin lung, prostate, colorectal, breast, spleen, liver, and colon cancer. ‘Integrin
binding’ was induced in lung, renal, breast, spleen, liver, and colon cancer. ‘Metalloendopeptidase inhibitor
activity’, metalloenzyme regulator, ‘Metalloenzyme inhibitor’ and ‘Insulin-like growth factor binding’ was
IAEES www.iaees.org
6 Network Biology, 2014, 4(1): 1-20

inhibited in most of the cancers including lung, prostate, colorectal, renal, spleen, liver, and colon cancer. The
pathways of ‘Metalloendopeptidase inhibitor activity’, metalloenzyme regulator, ‘metalloenzyme
inhibitor’were induced in prostate, colorectal, liver, and colon cancer, however, they were inhibited in lung,
renal, leukemia, breast, and spleen cancer. The GO term ‘Transaminase activity’was found induced in six
cancer namely prostate, colorectal, spleen, liver and colon cancer. Moreover, ‘Transforming growth factor 2
beta receptor activity (TGFB2)’ was inhibited in five cancers namely lung, prostate, colorectal, renal, and
spleen cancer (Fig. 2).
Fig.3 represents the up regulated molecular functions, which were shared among cancer at different sites.
‘Ubiquitin thiolesterase activity’ was observed in seven different cancers namely leukemia, breast,bladder,
colorectal, lung, prostate, and spleen cancer. ‘Phosphoserine phosphatase’ was identified in only three
cancers; leukemia, colon, and liver cancer. Fig. 4 depicts the down-regulated molecular functions among all
cancers. ‘Transforming growth factor beta receptor activity’was the most frequently appearing down
regulated molecular function, observed in breast, bladder, colorectal, lung, kidney, prostate, and spleen
cancer.

Fig. 3 Graph represents the molecular functions, which were up regulated in different cancers. On y-axis, biological terms
aredisplayed and x-axis represents the expression of molecular function in various cancers distinguished by different colors.
Legends displayed on the right side of the graph represent the list of cancers. ‘Ubiquitin thiolesterase activity’ is up regulated in
seven different cancers and on the other hand ‘phosphoserine phosphatase activity’ is up regulated in only threedifferent
cancers.

IAEES www.iaees.org
 Network Biology, 2014, 4(1): 1-20 7

Fig. 4 Graph represents the molecular functions, which were down regulated in different cancers. On y-axis, biological terms
aredisplayed and x-axis represents the expression of molecular function in various cancers distinguished by different colors.
Legends displayed on the right side of the graph represent the list of cancers. Down regulation of ‘Transforming growth factor
beta receptor activity’ is frequently appearing molecular function observed among seven different cancers.

The prediction of over represented TF binding sites (TFBS) and transcription factors (TF) in different
metastatic cancers was accomplished by using oPOSSUM. A complete list of TFBS for all the cancers is
provided in file S2. Table 2 enlists the top ranked TFBS in all the cancer with maximum Z scores. Table 3
gives an overview of all TFBS that were shared (appeared in more than one cancer) and/or appeared uniquely
(that were tissue specific and found only in one cancer) in all cancers.
We analyzed the list of DEGs for all cancerous data sets with oPOSSUM and found that the binding sites
of Broad complex-3 and Broad complex-4, SRY were indeed the most significantly over-represented (Table
2 and 3). Broad complex-3 and Broad complex-4, SRY sites were identified in most of the target gene in all
the cancer. Broad complex 3 and Broad complex 4 were present in all the cancers except, leukemia and
bladder cancer. Hb, MYB.ph3, NKx 2-5 and Sox 2 were shared among seven cancers while Dof2 and Prrx2
were shared among five cancers. PBF was found unique for colorectal cancer, FOX 11 for renal cancer, and
NF-kappaB for leukemia. Agamous, AGL3, Hand1-Tcfe2a, SP1 and TBP were listed only in bladder cancer
(Table 3).
Network analysis was done by using Cytoscape (v2.8.2). Genes were mapped against their respective TF,

IAEES www.iaees.org
8 Network Biology, 2014, 4(1): 1-20

represented by two different colors i.e. yellow color for TFsand pink color for genes expressed in different
cancers. In network analysis, 100 Gene regulatory networks between TF and genes were generated. We have
displayed the network of leukemia, which shows an apparently different result from other cancers, and a
network of SRY (spleen cancer). It is a common TF observed among the majority of cancers (Fig. 5-6).

Table 2 Statistically over-represented TF binding sites in gene expression data sets of different cancers.
TF TF Class Z-score Fisher Score TF TF Class Z-score Fisher Score
A: Lung Cancer F: Spleen Cancer

Broad-complex_3 ZN-FINGER, 51.8 4.66E-17 Broad-complex_3 ZN-FINGER, 36.59 1.1E-10

C2H2 C2H2
Nkx2-5 HOMEO 49.56 5.14E-08 SRY HMG 33.25 0.0000003
SRY HMG 49.42 2.71E-09 Sox5 HMG 32.93 1.55E-08
Broad-complex_4 ZN-FINGER, 46.51 2.55E-17 Broad-complex_4 ZN-FINGER, 32.58 1.65E-11
C2H2 _ C2H2
B: Prostate Cancer G: Liver Cancer

Broad-complex_3 ZN-FINGER, 41.67 1.47E-11 Broad-complex_3 ZN-FINGER, 36.98 6.87E-14

C2H2 C2H2
Nkx2-5 HOMEO 37.63 0.000011 Sox5 HMG 33.48 1.56E-09
SRY HMG 37.51 0.000004 SRY HMG 32.77 1.52E-08
Broad-complex_4 ZN-FINGER 37.28 3.88E-11 Broad-complex_4 ZN-FINGER 32.29 3.63E-13
C2H2 C2H2
C: Colorectal Cancer H: Colon Cancer

Broad-complex_3 ZN-FINGER, 12.98 0.01 Broad-complex_3 ZN-FINGER, 44.69 2.15E-14

C2H2 C2H2
SRF MADS 11.84 0.0118 Broad-complex_4 ZN-FINGER 40.12 6.27E-15
C2H2
NR3C1 NUCLEAR 9.665 0.000568 SRY HMG 39.56 2.11E-08
RECEPTOR
MYB.ph3 TRP-CLUS- 9.284 0.00396 Nkx2-5 HOMEO 39.18 0.000001
TER
D: PPRC I: Leukemia

SRY HMG 28.69 0.00000185 RREB1 ZN-FINGER 12.43 0.0118

Broad-complex_3 ZN-FINGER 26.43 2.38E-09 ELF5 ETS 9.968 0.00459
C2H2
Nkx2-5 HOMEO 25.06 0.0000118 RELA REL 8.161 0.00008
Broad-complex_4 ZN-FINGER 24.04 8.23E-09 REL REL 8.064 0.00931
C2H2
E: Breast Cancer J: Bladder Cancer

Broad-complex_3 ZN-FINGER, 32.5 1.35E-09 SRF MADS 22.4 3.56E-10

C2H2
Sox5 HMG 28.06 0.0000011 FOXD1 FORKHEAD 18.69 3.37E-08
Broad-complex_4 ZN-FINGER 26.54 2.41E-09 TBP TATA-box 17.83 0.000013
C2H2
SRY HMG 25.66 0.0000136 SRY HMG 16.32 0.0000356
TFBSs detected by oPOSSUM with the top 4 mostly highly ranked Z-scores or with Fisher P-value < 0.01. TFs over-expressed or
inhibited in gene expression studies.

IAEES www.iaees.org
 Network Biology, 2014, 4(1): 1-20 9

Table 3 Summary of over-represented TF binding sites in gene expression data sets of all cancers.

List of TF Lung Prostate Colorectal Renal Breast Spleen Liver Colon Leukemia Bladder
Agamous - - - - - - - - - Yes
AGL3 - - - - - - - - - Yes
Broad- Yes Yes Yes Yes Yes Yes Yes Yes - -
Complex_3
Broad Yes Yes Yes Yes Yes Yes Yes Yes - -
Complex _4
Dl_1 - - - - - - - - Yes -
Dl_2 - - - - - - - - Yes -
Dof2 - Yes Yes - Yes Yes Yes - - -
Dof3 - - - - Yes Yes Yes - - -
ELF5 - - - - - - - - Yes -
FOXD1 - - - - - - - - Yes Yes
Foxd3 Yes Yes Yes Yes Yes Yes Yes Yes - -
FOXF2 - - Yes - - - - - - Yes
FOXI1 - - - Yes - - - - - -
Hand1-Tcfe2a - - - - - - - - - Yes
Hb Yes Yes - Yes Yes Yes Yes Yes - -
MYB.ph3 Yes Yes Yes - Yes Yes Yes Yes - -
NF-kappaB - - - - - - - - Yes -
Nkx2-5 Yes Yes - Yes Yes Yes Yes Yes - -
NR3C1 - - Yes - - - - - - -
PBF - - Yes - - - - - - -
Pdx1 Yes - - Yes - - - - - -
Prrx2 Yes Yes Yes Yes - - - Yes - -
REL - - - - - - - - Yes -
RELA - - - - - - - - Yes -
RREB1 - - - - - - - - Yes -
RUNX1 - - - - - - - - Yes -
Sox5 Yes Yes - Yes Yes Yes Yes Yes - -
SP1 - - - - - - - - - Yes
SQUA - - - - - - - Yes - Yes
SRF - - Yes - - - - - Yes Yes
SRY Yes Yes - Yes Yes Yes Yes Yes - Yes
TBP - - - - - - - - - Yes

Set of genes in the included target set from which TFBS was identified is listed under Target gene hits
Z-score is the possibility that the number of TFBS nucleotides predicted in the included target genes is significant as compared to the
TFBSnucleotides predicted in the Background set.
Fisher Score is the possibility that the number of hits vs. non-hits in the included target genes could have occurred in random chance based
on the number of hits vs. non-hits of the background.

IAEES www.iaees.org
10 Network Biology, 2014, 4(1): 1-20

A B

C D

Fig. 5 A, B, C and D networks of leukemia representing association of transcription factors SRF, dl_1, ELF5 and
FOXD1(yellow color) with its respective genes (pink color), generated by using Cytoscape (v. 2.0)

4 Discussion
According to the objectives of our study, we primarily analyzed the microarray data sets of metastatic cancers
at different sites to visualize the most impacted biological terms. The analysis revealed not only a set of the
most impacted biological terms in all selected cancers, but also uncovered a list of TF and their biological
networks in the datasets. The data further uncovered the biological terms that were shared by most of the
cancers as well as the set of genes that were unique to a particular tissue.
Extracellular matrix (ECM) not only contributes to the structural integrity of the tissues but also mediates
the signal transduction processes that conduct growth, differentiation and cell migration. Expression and
activation of some tumor related genes is known be associated directly with ECM (Adams and Watt, 1993;

IAEES www.iaees.org
 Network Biology, 2014, 4(1): 1-20 11

Howlett et al., 1994). The exact mechanism of ECM action is still not known, but deregulation of ECM
promotes metastasis by destabilizing the cell structural organization and basement membrane that allow the
passage of tumor cell. As the result of such degradation, primary tumor proliferates and penetrates into new
tissue habitats (Nerenberg et al., 2007). Growth kinetics of the cell highly depends on the type of growth
factor (GF) that binds the ECM and its microenvironment (Petersenet al., 1992). Upregulation of ECM
constituents in current study might have promoted the specific set of GFs that might have positively
influenced the proliferation and cell invasion. Furthermore, collagen is the primary constituent of
physiological barriers, and it has assumed to be the primary ground upon which biochemical events of
metastasis take place (Nerenberg et al., 2007). Verhoeven et al. (1993) have reported an increased expression
of type IV collagen in periductal breast cancer. The induction of collagen binding in most of the cancers might
be indicative of the fact that important physiological changes occur in non-basement membrane of ECM.

A B

C D

Fig. 6 A, B, C and D networks of spleen cancer generated by using Cytoscape (v. 2.0),represents association of
transcriptionfactors SRY, Broad-complex_3, Broad-complex_4 and Foxd3 (yellow color) with its respective genes (pink color).

IAEES www.iaees.org
12 Network Biology, 2014, 4(1): 1-20

The function of metalloendopeptidase inhibitor is to inhibit the matrix metalloendopeptidase (MMP) that
is associated with the extracellular matrix (ECM) turn over. Metalloenzyme regulator and metalloenzyme
inhibitor are the proteins that regulate and inhibit the metal containing enzymes that can modulate many
diseases including cancer. These inhibitors play an important role in the homeostasis of ECM, which is a
crucial player of tumor invasion and metastasis (Jiang et al., 2002). Deregulation of the metallo-
endopeptidase inhibitor activity is reported to be associated with migration of the tumor cell and
neovascularization. Prolyloligopeptidase activity is elevated in various cancers and plays an important role in
the promotion of angiogenesis (Christiansen et al., 2013). Transminase activity is the catalyst for stimulating
the transfer of amino group to acceptor, an aid to detection of liver metastases (Kim et al., 1977).
Insulin-like growth factor (IGF) binding promotes cell migration and has been shown to either inhibit or
stimulate growth-promoting effects of Insulin-like Growth Factors (IGFs). They alter the interaction of IGFs
with their cell surface receptors that triggers the cascade of molecular events, and ultimately lead to
malignancy (Sala et al., 2005; Samani et al., 2007). Inhibition of IGF binding in most of the cancers in our
study dictates a diverse role of this pathway in controlling the cell migration and abnormal proliferation
pathways. Transforming growth factor 2 beta receptor activity (TGFB2) regulates a plethora of physiological
and pathological processes such as cell cycle arrest in epithelial and hematopoietic cells, control of
mesenchymal cell proliferation and differentiation, wound healing extracellular matrix production
immunosuppression and carcinogenesis. TGFB2 plays an important role in tumor invasion and metastasis
(Ganapathy et al., 2010). Microtubule plus end binding is known as the most crucial player in many cellular
functions, such as cell migration. The regulation of cell migration is more activated in cancer cells when
cancer cells become metastatic (Scolz et al., 2012).
Phosphoserine phosphatase activity plays a fundamental role in the regulation of a number of signaling
pathways whose deregulation can contribute to cancer (Eichhorn et al., 2009). The most important pathways
include Notch, Wnt and Hedgehog. Notch signaling is the key regulator of development and cell fate in
various tissues. Notch signaling activates proto-oncogenes; Notch 1 and -3, they play a crucial role in
carcinogenesis. Similarly, Wnt signaling produces a proto-oncogene Wnt1 that stimulates oncogenesis.
Hedgehog signaling is mostly identified in basal cell carcinomas (Malhotra et al., 2011). Threonine type
endopeptidase activity has associated pathological processes such as carcinogenesis. This endopeptidase
activity through its upregulation plays a central role in tumor growth and the multistep processes of invasion
and metastasis, as it destabilize the cell stability (Gialeli et al., 2011; Yu et al., 2010). These evidence
supports the upregulation of phosphoserine phosphatase activity in many cancers in this study. Upregulation
of phosphoserine phosphatase activity, threonine type endopeptidase activity in conjunction with down
regulation of TGF beta-receptor activity that is a tumor suppressor gene, are combined factors that lead to
malignancy.
Ubiquitin thiolesterase activity is involved in the enhancement of proliferation in multiple types of cells.
Kabuta et al has probed this activity by using various mutants of ubiquitin thiolesterase. Inhibition of the
interaction between ubiquitin thiolesterase activity and cell cycle-associated CDK resulted in the destruction
of ubiquitin thiolesterase-induced enhancement of cell proliferation (Kabuta et al., 2013). Ubiquitin
thiolesterase activity is observed in the majority of cancers in our study, which emphasizes its important role
in boosting cellular proliferation. Besides this, phosphoserine phosphatase activity, although identified as up
regulated in few cancers in our study, because it plays a discrete role in signals pathway regulation, also has
tumor suppressor characteristic.
Cadherins are calcium-dependent cell adhesion proteins that trigger loss of cell–cell adhesion; cadherin
might also deliver signals that actively induce tumor-cell invasion and metastasis (Vleminckx et al., 1991).

IAEES www.iaees.org
 Network Biology, 2014, 4(1): 1-20 13

Cadherin binding that was observed as an up regulated function in bladder, colon, breast, kidney, prostate
cancer, is pivotal in maintaining epithelial tissue integrity and are the major barrier for epithelial cancer
metastasis (Imai et al., 2000; Lu, 2010). It has been observed as a key element of tumor progression in many
studies, serving as a suppressor of invasion and metastasis (Hajra and Fearon, 2002). Cadherin binding is also
the important molecular function observed in many cancers in our study as it actively induces and supports
tumor development.
Major histocompatibility complex I (MHC) Class I and MHC protein binding presented the same
response in many cancers in our study. These are well known to trigger the immune system in fighting
against cancerous cells. MHC Class I protein binding is involved in the presentation of foreign antigens to the
immune system (Watkins et al., 1990). MHC Class I A is commonly expressed in carcinoma cell lines. It is
phenotypically detected mostly in T cell leukemia cell lines. Proper functioning of MHC I molecules are
essential in fighting cancer (Evans et al., 1999) for recognition of cancerous cell in order to generate a proper
immune response (Pende et al., 2002). MHC protein binding plays a role in the natural control of cancer cells.
Cancer cells contain many mutated proteins that may be presented by MHC to alert the immune system.
Tumor cells may also express normal proteins but in rare places or in abnormal amounts, providing a
potential signal to mobilize an immune response (Goodsell, 2005).
snRNA-binding is a molecular function that plays different yet an important role. It plays a role in
aggravating cancer by overexpression of genes during splicing. snRNA binding is among the increasing list
of splicing factors that have been found to be upregulated or downregulated in cancers, as compared with the
equivalent normal tissues. Cancer-associated genetic instability is likely to have an important role in this
process. Overexpression of splicing factor like SF2/ASF was reported to associate with amplification of the
gene encoding these splicing factors, thus, enhancing the genetic instability (Karni et al., 2007), whereas
reduced expression of many genes like RBM5 in lung cancer correlates with deletion of its gene locus at
chromosomal region 3p21.3 (Oh et al., 2006).
Structural constituent of ribosome exhibits its role in biological processes including cell adhesion,
differentiation, migration, signaling, neurite outgrowth and metastasis (Hao et al., 2004). Ribosomal proteins
control cell cycle functioning such as checkpoints of cell cycle and cell proliferation. They play a crucial role
in translational regulation and control of cellular transformation, tumor growth, aggressiveness and metastasis
(Chattopadhyay et al., 2007).
Water channel activity includes specific membrane proteins that play important roles in numerous
physiological processes such as adhesion and migration. Studies indicated that abnormal expression or
activity of a number of ion channels e.g. voltage-gated K+, Na+, Ca2+ channels, TRP channels, and
epithelial Na+/degenerin family of ion channels, are involved in the growth/proliferation, migration or
invasion of cancer cells (Li and Xiong, 2011). In our study, water channel activity is found down regulated in
many cancers as it enhances abnormal cell growth and expression.
Vascular endothelial growth factor (VEGF) is associated with tumor angiogenesis and poor prognosis in
human colorectal cancer. VEGF receptor-1 is a high-affinity receptor for VEGF and is typically considered
specific to endothelial cells. VEGFR-1 was expressed in all colorectal cell lines. Both VEGF-A and -B led to
significant induction of cell motility and invasiveness of colorectal cells. VEGF family ligands can activate
processes involved in tumor progression and metastasis (Fan et al., 2005). VEGF was observed to be
upregulated in many cancers, however, unexpected down regulation was observed in, colorectal, kidney, lung
and prostate cancer.
Transforming growth factor-β (transforming growth factor-β (TGF-β) has emerged as a family of growth
factors involved in essential physiological processes, including embryonic development, differentiation,

IAEES www.iaees.org
14 Network Biology, 2014, 4(1): 1-20

tissue repair and cell growth control (Javelaud and Mauviel, 2004). Alterations in the TGF-β signaling
pathway, including mutation or deletion of members of the signaling pathway and resistance to TGF-β-
mediated inhibition of proliferation are frequently observed in human cancers (Elliott and Blobe, 2005).
Down regulation in our study supports these findings. TGF is down regulated in bladder, colorectal,
leukemia, prostate and spleen cancer. The role of transforming growth factor beta (TGF-beta) in
carcinogenesis is quite complex, with tumor suppressor and pro-oncogenic activities depending on the
particular tumor cell and its stage in malignant progression. Smad2/3, important players in TGF signaling,
played a prominent role in regulating tumor suppressor effects on well-differentiated breast cancer cell lines
resulting in the formation of larger tumors with an increased proliferation and more malignant histologic
features (Tian et al., 2004). VGEF, TGF-beta and SH2 binding are the three important molecular functions
mentioned above were identified as down regulated in many cancers. This may be due to their almost similar
role in inhibiting cancer progression because all three highly promote proto-oncogenic activities that’s why
they need to be down regulated to maintain the balance in a cell.
Chemokines and their receptors play critical roles in leukocyte trafficking during inflammatory processes.
Although the role of chemokine receptors (CKRs) in cancer biology is relatively new, the data suggest that a
number of CKRs, including CXCR4, CCR4, CCR7, and CCR10, may play diverse roles in cancer growth,
metastasis, and angiogenesis, or the composition of the cancer microenvironment. For example, preclinical
models of cancer indicate that cancer antagonists, most notably those for CXCR4, can block cancer growth
either directly or by altering the cancer stroma (Wu et al., 2009). Similar to these observations, chemokines
activity is identified as down regulated in leukemia, bladder, liver, lung and kidney cancer in our study.
Src homology-2 (SH2) domain containing phosphatases (Shps) are small, highly conserved subfamily of
protein-tyrosine phosphatases. It has been identified that mutations in human Shp2 as the cause of the
inherited disorder Noonan syndrome. Shp2 mutations might also contribute to the pathogenesis of some
leukemia’s. These domains are the elements that control the interaction of cytoplasmic signaling proteins
(Neel et al., 2003). SH2 binding is down regulated in bladder, breast, lung, kidney, prostate, and spleen
cancer. In our study, the SH2 domain binding is observed as down regulated in many cancers because of role
in controlling the signaling pathways. But with reference to the study described above, our results for
leukemia were found to be different because in our study leukemia is not reported to be down regulated in
SH2 binding domain. This may be due to the difference in choosing the cancer type, pathological grade or
primary site in our study comparable to their study.
Phosphoinositide kinase 3 binding activity is down regulated in breast, lung, leukemia and spleen cancer.
It is crucial for cell invasion and migration, not only for physically tethering cells to the matrix, but also for
sending and receiving molecular signals that regulate these processes (Engelman, 2007). It establishes the
initiation of an intricate signaling cascade, which eventually results in the mediation of cellular activities such
as proliferation, differentiation, chemotaxis, survival, trafficking, and glucose homeostasis (Katso et al.,
2001). Broad-complex (BR-C_3 or 4) belongs to class of zinc finger transcription factors, observed in all
cancers except leukemia and bladder cancer. It is a primarily ecdysone response gene, which is a key
regulator of metamorphosis, which enhances cell growth and differentiation (Karim et al., 1993). On the
contrary, FoxD3 was observed in kidney, lung, prostate, colorectal, spleen, breast, liver and colon cancer.
FoxD3 belongs to the forkhead family of transcription factors, acts as transcription repressor and activator.
Important player required for the maintenance of pluripotent cells in the various embryogenesis stages as it
inhibits migration and invasion of cells (Guo et al., 2002).
In conjunction with the other shared transcription factors, SRY is a member of transcription factor class
HMG. SRY is a transcriptional regulator that controls a genetic switch in male development. SRY is also

IAEES www.iaees.org
 Network Biology, 2014, 4(1): 1-20 15

involved in different aspects of gene regulation including activation and repression of tumor. It plays an
important role in pre-mRNA splicing. Elevated expression of SRY-related HMG box transcription factor has
been reported in various tumors, comprising colorectal cancer, lung cancer, and breast cancer. SRY plays a
central role in the development of these malignancies (Vervoort et al., 2013).
ELF5, NF-kappa B, Dl_1, REL, RELA, RREB1 and RUNX1 are predicted only in leukemia. L_1 and
dl_1 both belongs to REL family of transcription factors. ELF5 belongs to epithelium specific subclass of the
ETS transcription factor family (Zhou et al., 1998). ELF5 regulates the later stages of terminal differentiation
of keratinocytes and regulates a number of epithelium-specific genes found in tissue containing glandular
epithelium such as salivary glands and prostate (Sharrocks et al., 1997). Likewise, NF-kappa B, REL, RELA
all belong to REL family, it is a pleiotropic transcription factor which is present in almost all cell types and is
involved in various biological process such as inflammation, immunity, differentiation, cell growth
tumorigenesis and apoptosis. NF-kappa B is controlled by various mechanisms of post-translational
modification and sub cellular compartmentalization as well as by interactions with other co-factors or co-
repressors (Chen et al., 1999). Correspondingly, RREB1 belongs to ZN-FINGER, C2H2 family. RREB1
binds specifically to the RAS-responsive elements (RRE) of gene promoters it may be involved in Ras/Raf-
mediated cell differentiation by enhancing calcitonin expression and this further leads to the medullary
thyroid carcinoma (MTC) (Carson et al., 1995). RREB1 represses the angiotensinogen gene. It negatively
regulates the transcription activity of androgen receptor Stimulates the transcriptional activity of NEUROD1.
RUNX1 belongs to RUNT family. It is a heterodimeric transcription factor that binds to the core element of
many enhancers and promoters (Zhang et al., 1997). The protein encoded by RUNX1 is involved in the
development of normal hematopoiesis. Chromosomal translocations involving this gene were well
documented and have been associated with several types of leukemia (Ichikawa et al., 2013).
TBP that was found only in leukemia belongs to TATA-Box family; it is a general transcription factor
that functions at the core of the DNA binding multi protein factors TFIIT. Binding of TFIIT to the TATA box
is the initial transcriptional step of the peri-initiation complex, playing a major role in the activation of
eukaryotic genes transcribed by RNA polymerase II (Peterson et al., 1990). Similarly, SP1 belongs to same
family of transcription factor as RREB1 i.e. ZN-FINGER, C2H2 family. SP1 can activate or repress
transcription in response to physiological and pathological stimuli. It binds with high affinity to GC-rich
motif and regulates the expression of large number of genes involved in a variety of process such as cell
growth, apoptosis, differentiation and immune response. SP1 highly regulated by post translational responses.
It may have a role in modulating the cellular response to DNA damage. TBP and SP1 both show expression
in regulating cellular proliferation and apoptosis (Wimmer et al., 1997).
NR3C1 that appeared only in colorectal cancer belongs to ZN-FINGER and DOF family. NR3C1 gene
encodes glucocorticoid response elements in the promoters of glucocorticoid responsive genes to activate
their transcription, and as a regulator of other transcription factors. This receptor is typically found in
cytoplasm, but upon ligand binding is transported in to the nucleus. It is involved in inflammatory responses,
cell proliferation and differentiation in target tissues (Ray et al., 1996).
In papillary renal cell carcinoma, FOXI1 gene appears unique. FOXI1 belongs to FORKHEAD family of
transcription factors, which is characterized by a distinct FORKHEAD domain. FOXI1 plays a major role in
the development of cochlea and vestibulum, as well as embryogenesis (Larsson et al., 1995). Deregulation of
FOX family transcription factors can alter the cell fate, boost cancer development and progression (Myatt and
Lam, 2007). FORKHEAD transcription factors are the crucial player in the regulation of cell cycle arrest, cell
death and DNA damage repair (Yang and Hung, 2011).

IAEES www.iaees.org
16 Network Biology, 2014, 4(1): 1-20

5 Conclusions
The use of bioinformatics techniques for a holistic understanding of any biological system needs to be
addressed with all the biological elements involved in a system like transcriptome, proteome and their
regulator networks. The use of system biology to study cancer and its pathological effects needs to go beyond
simple functional analysis of transcriptome. Instability and invasiveness of the cancers reported in this study
were clearly depicted by deregulation of pathways like extracellular matrix structural constituent, metallo-
endopeptidase inhibitor, insulin-like growth factor (IGF), phosphoserine phosphatase activity and cadherins
at the transcriptome level. However, to confirm and specify the role of these pathways in the deadly spread of
cancer, regulator pattern of these pathways have to be identified at protein level as well to screen candidate
for drug targets. This will bring a deeper understanding of complex microenvironment of cancer with the
improvement in drug candidates against cancer.

References
Auguste P, Fallavollita L, Wang N, et al. 2007. The host inflammatory response promotes liver metastasis
byincreasing tumor cell arrest and extravasation. American Journal of Pathology, 170(5): 1781-1792
Adams JC, Watt FM. 1993.Regulation of development and differentiation by the extracellular
matrix.Development, 117(4): 1183-1198
Bardelli A, Saha S, Sager JA, et al. 2003. PRL-3 expression in metastatic cancers. Clinical Cancer Research,
9(15): 5607-5615
Bionaz M, Periasamy K, Rodriguez-Zas SL, et al. 2012. A novel dynamic impact approach (DIA) for
functional analysis of time-course omics studies: validation using the bovine mammary transcriptome.
PLoS One, 7(3): e32455
Blancafort P, Segal DJ, Barbas, CF. 2004. Designing transcription factor architectures for drug discovery.
Molecular Pharmacology, 66(6): 1361-1371
Carson EB, McMahon M, Baylin SB, et al. 1995. Ret gene silencing is associated with Raf-1-induced
medullary thyroid carcinoma cell differentiation. Cancer Research, 55(10): 2048-2052
Chattopadhyay I, Kapur S, Purkayastha J, et al. 2007.Gene expression profile of esophageal cancer in North
East India by cDNA microarray analysis. World Journal of Gastroenterology, 13(9): 1438-1444
Chen F, Castranova V, Shi X, et al. 1999. New insights into the role of nuclear factor-kappaB, a ubiquitous
transcription factor in the initiation of diseases. Clinical Chemistry, 45(1): 7-17
Christiansen V J, Jackson KW, Lee KN, et al. 2013.Targeting inhibition of fibroblast activation protein-alpha
and prolyloligopeptidase activities on cells common to metastatic tumor microenvironments.Neoplasia,
15(4): 348-358
Dunker AK, Uversky VN, et al. 2010. Drugs for 'protein clouds': targeting intrinsically disordered
transcription factors. Current Opinion in Pharmacology, 10(6): 782-788
Edgar R, Barrett T, et al. 2006. NCBI GEO standards and services for microarray data. Nature Biotechnology,
24(12): 1471-1472
EichhornPJ, Creyghton MP, Bernards R, et al. 2009. Protein phosphatase 2A regulatory subunits and
cancer.BiochimicaetBiophysicaActa, 1795(1): 1-15
Elliott RL, Blobe GC. 2005. Role of transforming growth factor Beta in human cancer. Journal Clinical
Oncology, 23(9): 2078-2093

IAEES www.iaees.org
 Network Biology, 2014, 4(1): 1-20 17

Emilsson V, Thorleifsson G, Zhang B et al. 2008. Genetics of gene expression and its effect on disease.
Nature, 452(7186): 423-428
Engelman JA. 2007. The role of phosphoinositide 3-kinase pathway inhibitors in the treatment of lung cancer.
Clinical Cancer Research, 13(15 Pt 2): s4637-4640
Evans DT, Knapp LA, Jing P, et al. 1999. Three different MHC class I molecules bind the same CTL epitope
of the influenza virus in a primate species with limited MHC class I diversity. Journal of Immunology,
162(7): 3970-3977
Fan F, Wey JS, McCarty MF, et al. 2005. Expression and function of vascular endothelial growth factor
receptor-1 on human colorectal cancer cells. Oncogene, 24(16): 2647-2653
Ganapathy V, Ge R, Grazioli A, et al. 2010. Targeting the Transforming Growth Factor-beta pathway inhibits
human basal-like breast cancer metastasis. Molecular Cancer, 9: 122
Gialeli C, Theocharis AD, KaramanosNK. 2011. Roles of matrix metalloproteinases in cancer progression
and their pharmacological targeting. FEBS Journal, 278(1): 16-27
Goodsell DS. 2005. The molecular perspective: major histocompatibility complex. Oncologist, 10(1): 80-81
Guo Y, Costa R, Ramsey H, et al. 2002. The embryonic stem cell transcription factors Oct-4 and FoxD3
interact to regulate endodermal-specific promoter expression. Proceedings of the National Academy of
Sciences USA, 99(6): 3663-3667
Hajra KM, Fearon ER. 2002. Cadherin and catenin alterations in human cancer. Genes Chromosomes Cancer,
34(3): 255-268
Hao X, Sun B, Hu L, et al. 2004.Differential gene and protein expression in primary breast malignancies and
their lymph node metastases as revealed by combined cDNA microarray and tissue microarray analysis.
Cancer, 100(6): 1110-1122
Howlett AR, Petersen OW, Steeg PS, et al. 1994. A novel function for the nm23-H1 gene: overexpression in
human breast carcinoma cells leads to the formation of basement membrane and growth arrest. Journal of
National Cancer Institute, 86(24): 1838-1844
Huang S, Fulton D, et al. 2006.Identification of over-represented combinations of transcription factor binding
sites in sets of co-expressed genes. In: Advances in Bioinformatics and Computational Biology (Vol. 3).
247-256, Imperial College Press, London, UK
Ichikawa M, Yoshimi A, Nakagawa M, et al. 2013.A role for RUNX1 in hematopoiesis and myeloid
leukemia. International Journal of Hematology, 97(6): 726-734.
Imai K, Takada N, Satoh N, et al. 2000. (Beta)-catenin mediates the specification of endoderm cells in
ascidian embryos. Development, 127(14): 3009-3020
Javelaud D, Mauviel A. 2004. Mammalian transforming growth factor-betas: Smad signaling and physio-
pathological roles. International Journal of Biochemistry and Cell Biology, 36(7): 1161-1165
JiangY, Goldberg ID, Shi YE. 2002. Complex roles of tissue inhibitors of metalloproteinases in cancer.
Oncogene, 21(14): 2245-2252
Kabuta T, Mitsui T, Takahashi M, et al. 2013. Ubiquitin C-terminal hydrolase L1 (UCH-L1) acts as a novel
potentiator of cyclin-dependent kinases to enhance cell proliferation independently of its hydrolase
activity. Journal of Biology and Chemistry, 288(18): 12615-12626
Karim FD, Guild GM, Thummel CS. 1993. The Drosophila Broad-Complex plays a key role in controlling
ecdysone-regulated gene expression at the onset of metamorphosis. Development, 118(3): 977-988
Karni R, de Stanchina E, Lowe SW, et al. 2007. The gene encoding the splicing factor SF2/ASF is a proto-
oncogene. Nature Structural & Molecular Biology, 14(3): 185-193

IAEES www.iaees.org
18 Network Biology, 2014, 4(1): 1-20

Katso R, Okkenhaug K, Ahmadi K, et al. 2001. Cellular function of phosphoinositide 3-kinases: implications
for development, homeostasis, and cancer. Annual Review of Cell and Developmental Biology, 17: 615-
675
Kim NK, YasminehWG, FreierEF, et al. 1977. Value of alkaline phosphatase, 5'-nucleotidase, gamma-
glutamyltransferase, and glutamate dehydrogenase activity measurements (single and combined) in serum
in diagnosis of metastasis to the liver. Clinical Chemistry, 23(11): 2034-2038
Kotelnikova E, Shkrob MA, Pyatnitskiy MA, et al. 2012. Novel approach to meta-analysis of microarray
datasets reveals muscle remodeling-related drug targets and biomarkers in Duchenne muscular dystrophy.
PLoS Computational Biology, 8(2): e1002365
Larsson C, Hellqvist M, Pierrou S, et al. 1995. Chromosomal localization of six human forkhead genes,
freac-1 (FKHL5), -3 (FKHL7), -4 (FKHL8), -5 (FKHL9), -6 (FKHL10), and -8 (FKHL12). Genomics,
30(3): 464-469
Li M, XiongZG. 2011. Ion channels as targets for cancer therapy. International Journal of Physiology,
Pathophysiology and Pharmacology, 3(2): 156-166
Lopes CT, Franz M, Kazi F, et al. 2010. Cytoscape Web: an interactive web-based network browser.
Bioinformatics, 26(18): 2347-2348
Lu Q. 2010. delta-Catenin dys-regulation in cancer: interactions with E-cadherin and beyond. Journal of
Pathology, 222(2): 119-123
MalhotraGK, Zhao X, Band H, et al. 2011. Shared signaling pathways in normal and breast cancer stem cells.
Journal of Carcinogenesis, 10: 38
MoothaVK, Lindgren CM, Eriksson KF, et al. 2003. PGC-1alpha-responsive genes involved in oxidative
phosphorylation are coordinately downregulated in human diabetes. Nature Genetics, 34(3): 267-273
Myatt SS, Lam EW. 2007. The emerging roles of forkhead box (Fox) proteins in cancer. Nature Review of
Cancer, 7(11): 847-859
Neel BG, Gu H, PaoL. 2003. The 'Shp'ing news: SH2 domain-containing tyrosine phosphatases in cell
signaling. Trends in Biochemical Sciences, 28(6): 284-293
Nerenberg PS, Salsas-Escat R, Stultz CM. 2007. Collagen--a necessary accomplice in the metastatic process.
Cancer Genomics Proteomics, 4(5): 319-328
Nguyen DX, Massague J. 2007. Genetic determinants of cancer metastasis. Nature Reviews Genetics, 8(5):
341-352
Oh JJ, Razfar A, Delgado I, Reed R A, et al. 2006. 3p21.3 tumor suppressor gene H37/Luca15/RBM5
inhibits growth of human lung cancer cells through cell cycle arrest and apoptosis. Cancer Research,
66(7): 3419-3427
Pende D, Rivera P, Marcenaro S, et al. 2002. Major histocompatibility complex class I-related chain A and
UL16-binding protein expression on tumor cell lines of different histotypes: analysis of tumor
susceptibility to NKG2D-dependent natural killer cell cytotoxicity. Cancer Research, 62(21): 6178-6186
Petersen OW, Ronnov-Jessen L, Howlett AR, et al. 1992. Interaction with basement membrane serves to
rapidly distinguish growth and differentiation pattern of normal and malignant human breast epithelial
cells. Proceedings of the National Academy of Sciences of USA, 89(19): 9064-9068
Peterson MG, Tanese N, Pugh BF. 1990. Functional domains and upstream activation properties of cloned
human TATA binding protein. Science, 248(4963): 1625-1630
Ray DW, Davis JR, White A, et al. 1996. Glucocorticoid receptor structure and function in glucocorticoid-
resistant small cell lung carcinoma cells. Cancer Research, 56(14): 3276-3280

IAEES www.iaees.org
 Network Biology, 2014, 4(1): 1-20 19

Rhodes DR, Yu J, Shanker K, et al. 2004. Large-scale meta-analysis of cancer microarray data identifies
common transcriptional profiles of neoplastic transformation and progression. Proceedings of the
National Academy of Sciences of USA, 101(25): 9309-9314
Sala A, Capaldi S, Campagnoli M, et al. 2005. Structure and properties of the C-terminal domain of insulin-
like growth factor-binding protein-1 isolated from human amniotic fluid. Journal of Biological
Chemistry, 280(33): 29812-29819
Savli H, Sunnetci D, Cine N, et al. 2012. Gene expression profiling of B-CLL in Ukrainian patients in post-
Chernobyl period. Experimental Oncology, 34: 57-63
Samani AA, Yakar S, LeRoith D, et al. 2007. The role of the IGF system in cancer growth and metastasis:
overview and recent insights. Endocrine Reviews, 28(1): 20-47
Schadt EE, Lamb J, Yang X, et al. 2005. An integrative genomics approach to infer causal associations
between gene expression and disease. Nature Genetics, 37(7): 710-717
Schadt EE, Monks SA, Drake TA, et al. 2003. Genetics of gene expression surveyed in maize, mouse and
man. Nature, 422(6929): 297-302
Scolz M, Widlund PO, Piazza S, et al. 2012. GTSE1 is a microtubule plus-end tracking protein that regulates
EB1-dependent cell migration. PLoS One, 7(12): e51259
Sharrocks AD, Brown AL, Ling Y, et al. 1997. The ETS-domain transcription factor family.International
Journal of Biochemistry & Cell Biology, 29(12): 1371-1387
Smith LM, BirrerMJ. 1996. Use of transcription factors as agents and targets for drug development.
Oncology, 10(10): 1532-1538
Tian F, Byfield SD, Parks L, et al. 2004. Smad-binding defective mutant of transforming growth factor beta
type I receptor enhances tumorigenesis but suppresses metastasis of breast cancer cell lines. Cancer
Research, 64(13): 4523-4530
van 't Veer LJ, Dai H, van de VijverMJ, et al. 2002. Gene expression profiling predicts clinical outcome of
breast cancer. Nature, 415(6871): 530-536
Verhoeven D, Bourgeois N, Buyssens N, et al. 1993. Ultrastructural demonstration of type IV collagen
deposits in periductalelastosis in breast cancer. Pathology-Research and Practice, 189(2): 144-149
VervoortSJ, van Boxtel R, Coffer PJ. 2013. The role of SRY-related HMG box transcription factor 4 (SOX4)
in tumorigenesis and metastasis: friend or foe? Oncogene, 32(29): 3397-3409
Vleminckx K, Vakaet L, Jr. Mareel M, et al. 1991. Genetic manipulation of E-cadherin expression by
epithelial tumor cells reveals an invasion suppressor role. Cell, 66(1): 107-119
Wang K, Narayanan M, Zhong H, et al. 2009. Meta-analysis of inter-species liver co-expression networks
elucidates traits associated with common human diseases. PLoS Computational Biology, 5(12):
e1000616.
WaringJF, Gum R, MorfittD, et al. 2002. Identifying toxic mechanisms using DNA microarrays: evidence
that an experimental inhibitor of cell adhesion molecule expression signals through the aryl hydrocarbon
nuclear receptor. Toxicology, 181-182: 537-550
Watkins DI, Chen ZW, Hughes AL, et al. 1990. Evolution of the MHC class I genes of a New World primate
from ancestral homologues of human non-classical genes. Nature, 346(6279): 60-63
WimmerEA, Cohen SM, Jackle H, et al. 1997. Buttonhead does not contribute to a combinatorial code
proposed for Drosophila head development. Development, 124(8): 1509-1517
Wu X, Lee VC, Chevalier E, et al. 2009. Chemokine receptors as targets for cancer therapy.Current
Pharmaceutical Design, 15(7): 742-757
Yang JY, Hung MC. 2011. Deciphering the role of forkhead transcription factors in cancer therapy. Current
Drug Targets, 12(9): 1284-1290
IAEES www.iaees.org
20 Network Biology, 2014, 4(1): 1-20

Yu DM, Yao TW, Chowdhury S, et al. 2010.The dipeptidyl peptidase IV family in cancer and cell
biology.FEBS Journal, 277(5): 1126-1144
Zhang YW, Bae SC, Huang G, et al. 1997. A novel transcript encoding an N-terminally truncated
AML1/PEBP2 alphaB protein interferes with transactivation and blocks granulocytic differentiation of
32Dcl3 myeloid cells. Molecular and Cell Biology, 17(7): 4133-4145
Zhou J, Ng AY, TymmsMJ, et al. 1998. A novel transcription factor, ELF5, belongs to the ELF subfamily of
ETS genes and maps to human chromosome 11p13-15, a region subject to LOH and rearrangement in
human carcinoma cell lines. Oncogene, 17(21): 2719-2732
Zhu J, LumPY, Lamb J, et al. 2004. An integrative genomics approach to the reconstruction of gene networks
in segregating populations. Cytogenetic Genome Research, 105(2-4): 363-374
Zhu J, Jiang Z, Gao F, et al. 2011. A systematic analysis on DNA methylation and the expression of both
mRNA and microRNA in bladder cancer.PloS ONE, 6(11): e28223

IAEES www.iaees.org
Network Biology, 2014, 4(1): 21-30

Article

Local and global control of ecological and biological networks

Alessandro Ferrarini
Department of Evolutionary and Functional Biology, University of Parma, Via G. Saragat 4, I-43100 Parma, Italy
E-mail: sgtpm@libero.it, alessandro.ferrarini@unipr.it

Received 8 November 2013; Accepted 12 December 2013; Published online 1 March 2014

Abstract
Recently, I introduced a methodological framework so that ecological and biological networks can be
controlled both from inside and outside by coupling network dynamics and evolutionary modelling. The
endogenous control requires the network to be optimized at the beginning of its dynamics (by acting upon
nodes, edges or both) so that it will then go inertially to the desired state. Instead, the exogenous control
requires that exogenous controllers act upon the network at each time step. By the way, all my previous works
dealt with the goal of the global optimization of ecological and biological networks, i.e. how to drive them to a
desired final state. Here I face another pivotal question: how can we locally (step-by-step) drive ecological and
biological networks, so that also intermediate steps (not only the final state) are under our strict control? The
ratio behind this question is that intermediate dynamics could potentially go below or above critical
ecological-biological thresholds, hence invalidating the final global control. To this purpose, I introduce here a
modelling solution to the complete control of ecological and biological networks that couples local and global
control.

Keywords edges control; genetic algorithms; global dynamics; local dynamics; network control; nodes
control.

Network Biology     
ISSN 2220­8879   
URL: http://www.iaees.org/publications/journals/nb/online­version.asp 
RSS: http://www.iaees.org/publications/journals/nb/rss.xml 
E­mail: networkbiology@iaees.org 
Editor­in­Chief: WenJun Zhang 
Publisher: International Academy of Ecology and Environmental Sciences 

1 Introduction
Recently, I introduced a methodological framework so that ecological and biological networks (EBN from
now on) can be controlled (Ferrarini, 2011a). EBN can be efficaciously tamed from the outside (Ferrarini,
2013a) but also through the use of endogenous controllers (Ferrarini, 2013b), by coupling network dynamics
and evolutionary modelling. The endogenous control requires that the network is optimized at the beginning of
its dynamics (by acting upon nodes, edges or both) so that it will then go inertially to the desired state. Instead,
the exogenous control requires that exogenous controllers act upon the network at each cycle.
In another paper (Ferrarini, 2013c), I faced a further important question: how reliable is the achieved
solution? In other words, which is the degree of uncertainty about getting the desired result if values of edges

IAEES www.iaees.org
22 Network Biology, 2014, 4(1): 21-30

and nodes were a bit different from optimized ones? This is an important question, because it’s not assured that
while managing a certain system we are able to impose to nodes and edges exactly the optimized values we
would need in order to achieve the desired results. In order to face this topic, I have coined a 3-parts
framework (network dynamics - genetic optimization - stochastic simulations).
I have further introduced a solution to the choice of the most feasible solution to network control (Ferrarini,
2013d) by introducing the concepts of control success and feasibility.
By the way, all my previous works dealt with the goal of the global optimization of EBN, i.e. how to drive
them to a desired final state. Here I face another pivotal question: how can we locally (step-by-step) drive
ecological and biological networks, so that also intermediate steps (not only the final state) are under our
control? The ratio behind this question is that intermediate dynamics could potentially go below or above
critical ecological-biological thresholds, hence invalidating the final global control. To this purpose, I
introduce here a solution to the complete control of EBN by coupling local and global controls.

2 Mathematical Formulation
Given a generic ecological (or biological) dynamic system with n interacting actors

dS 
  (S , t ) (1)
dt
where Si is the amount (e.g., number of individuals, total biomass, covered surface etc…) of the generic i-th
actor, if we also consider inputs (e.g. species reintroductions) and outputs (e.g. hunting) from-to outside, we
must write:

dS   
  (S , t )  I  O (2)
dt
As noted by numerous authors (Liu et al., 2011; Slotine and Li, 1991) real systems’ dynamics of eq. (2) can be
modelled and simulated using a system of canonical, linear differential equations

 dS1
 dt  a11S1  ...  a1n Sn  I1  O1

... (3)
 dS
 n  an1S1  ...  ann S n  I n  On
 dt
that can be written in the compact form

dS   
 A* S  I  O (4)
dt
with initial values

S0 =<S1 (0), S2 (0)...Sn (0)> (5)

and co-domain limits

S1min  S1 (t )  S1max

... (6)
S
 n min  S n (t )  Sn max

IAEES www.iaees.org
 Network Biology, 2014, 4(1): 21-30 23

and where

 a11 ... a1n 

 
A   ... ... ...  (7)
a 
 n1 ... ann 
is the matrix of the per unit-of-time effect on Si due to unitary Sj.
Such systems can be tamed from outside using the following 1-external-controller model (Ferrarini,
2013a):
 dS1 
 dt   a11S1  ...  a1n S n  I1  O1  c11*C1*
  OPT

...

 dS n 
   an1S1  ...  ann S n  I n  On  cn1*C1* (8)
 dt  OPT
 dC1
  f1S1  ...  f n S n
 dt
where asterisks stand for the genetic optimization (Holland, 1975) of exogenous node’s edges (i.e., coefficients
of interaction with the inner system) and exogenous node’s stock, i.e. the modification of such values at the
beginning of network dynamics in order to get a certain goal (e.g., maximization of the final value of a certain
variable). In case 1 controller is not enough, the model in (8) must be expanded to the following
k-external-controllers model (Ferrarini, 2013a):

 dS1
 dt  a11S1  ...  a1n S n  I1  O1  c11*C1*  ...  c1k *Ck *

...
 dS
 n  an1S1  ...  ann S n  I n  On  cn1*C1*  ...  cnk *Ck *
 dt
 (9)
 dC1  f S  ...  f S
 dt 11 1 1n n

...

 dCk
 dt  f k 1S1  ...  f kn Sn
Alternatively, an ecological or biological network can be controlled from inside using the following control
model (Ferrarini, 2013b):

 dS1 
 dt   a11* S1*  ...  a1n* S n *  I1*  O1*
 OPT
...
 dS (10)
 n   an1* S1*  ...  ann* S n *  I n*  On*
 dt OPT

IAEES www.iaees.org
24 Network Biology, 2014, 4(1): 21-30

where asterisks stand for the optimization of edges (i.e., coefficients of interaction among variables) or nodes
(i.e., initial stocks), i.e. the modification of their values at the beginning of the network dynamics in order to
get a certain goal.
The control models in (8), (9) and (10) are able to globally drive any EBN to the desired final state with an
uncertainty degree that can be calculated as proposed in Ferrarini (2013c), and a feasibility degree as in
Ferrarini (2013d).
Now, let’s assume that we want to tame intermediate dynamics of EBN as well, not only the final state. By
introducing an intermediate control function (ICF), I conceive several possibilities.
a) we want the generic actor Si(t) to stay as close as possible to a certain value during its dynamics.
In this case, ICF must be as follow
E
ICF  
t 0
k  Si (t ) dt
(11)
where t is time (independent variable), E indicates the time at which the system goes to equilibrium, k is the
value at which we want Si(t) to be tied with, vertical lines indicates the module of the difference.
Now, we can reach our goal by optimizing our network using (8), (9) or (10) in order to achieve the
minimization of ICF.
As a result we can constrain Si(t) as close as possible to k along all its dynamics.
b) we want the generic actor Si(t) to stay as close as possible to a certain value but only during some steps of its
dynamics.
In the previous step, I have subdued each dynamical step of Si(t) from the beginning to the its final state. It
could be possible that we want to control Si(t) just in a certain time interval. In this case, eq. (11) becomes
m
ICF  
t n
k  Si (t ) dt
(12)
where n and m are generic points along the dynamic timeline. Again, eq. (12) must be minimized using (8), (9)
or (10) in order to achieve our goal.
c) we want to tie the generic actor Si(t) to a certain dynamic f(t) (e.g. exponential). In this case, we need the
following equation
E
ICF  
t 0
f (t )  Si (t ) dt
(13)
As a result, by minimizing ICF using (8), (9) or (10) we can constrain Si(t) as close as possible to the desired
dynamic function.
d) we want to tie the generic actor Si(t) to another network’s actor Sj(t). In this case, we need the following
equation to be minimized
E
ICF  
t 0
S j (t )  Si (t ) dt
(14)
e) we want to tie the generic actor Si(t) to a function of another network’s actor Sj(t). For instance we could
want Si(t) to be always twice as Sj(t) during network dynamics. In this case, we need the following equation to
be minimized using (8), (9) or (10)

IAEES www.iaees.org
 Network Biology, 2014, 4(1): 21-30 25

E
ICF  
t 0
f ( S j (t ))  Si (t ) dt
(15)
f) we want that the generic actor Si changes from step to step of a certain value u. To this aim, I use
E
dS (t )
ICF  
t 0
dt
- u dt
(16)
By minimizing ICF with (8), (9) or (10), we can impose that Si changes of exactly u at each time step.

3 An Applicative Example
Fig. 1 shows an ecological network borrowed with modification from Ferrarini (2012). Greenish nodes
represent positive actors (or events) for the goal of network control, i.e. the conservation of the target species
(in green at centre of the network). Reddish nodes represent ecological actors (or events) with negative impact
on the target species. Blueish nodes represent resources needed by the target species. The goal is to preserve
target species’ occurrence in the study area. Stocks stand for the actual amounts of individuals or biomass.
Updates (indicated as upd) stand for yearly internal dynamics (i.e., intraspecific gains due to births and/or
immigration rates minus losses due to deaths and/or emigration rates). Minimum and maximum values stand
for lowest and highest values of stock values. For the sake of simplicity, the maximum possible value for each
actor (in italic hereafter) has been set to 100. The percent value associated to links represents the percentage of
the receiver that is yearly consumed by the transmitter at the beginning of the network simulation. Road
mortality and re-introductions accounts for 18 and 10 individuals per year respectively (average values of the
last 5 years of field observations).

Fig. 1 The ecological network on which evolutionary control has been applied.

IAEES www.iaees.org
26 Network Biology, 2014, 4(1): 21-30

Since data are yearly-based, I have calculated inertial dynamics and optimized the above ecological system
using a system of difference recurrent equations, instead of differential ones.

Table 1 The interaction matrix relative to the ecological network of Fig. 1.

The previous ecological network has the following inertial dynamics (Fig. 2), with the target species
(green line) going extinction after 7 years.

Fig. 2 Resulting dynamics for the network of Fig. 1. X-axis measures time in years.

4 Local Network Control

Previous research revealed that the optimal solution may be to search at a high rate of crossover, a low rate of
mutation and proper population size (Kuo et al., 2000). In this study, crossover was set at a probability of 60%
while mutations occur with a probability of 5%. This low setting helps to avoid getting trapped local optima

IAEES www.iaees.org
 Network Biology, 2014, 4(1): 21-30 27

during the search (D’heygere et al., 2006). The initial population consisted of 5000 chromosomes that were
evolved over minimal 10,000 generations. I applied a steady-state genetic algorithm with a one-point crossover
operator (Wall, 1996) to accomplish crossover. In this case the parent genome strings are cut at some random
position to produce two “head” and two “tail” segments. The “tail” segments are swapped to produce two new
genomes. For parent selection the roulette wheel selection method was used (Goldberg, 1989), where the
likelihood of selection is proportionate to the fitness score given by the performance criterion. After crossover
and mutation, the individuals with the lowest fitness scores were removed.

Fig. 3 By using both local and a global optimizations, target species has been constrained close to the stock value 50 for 20 years
before going to its maximum.

Now let’s suppose we want to stabilize the target species to a steady value equal to 50 for at least 20 years
before it goes to its maximum. Let’s suppose we want to reach it by acting upon target species’ intraspecific
coefficient (actually equal to 1.30). To this aim, I used equations (10) and (12) and, being all other simulation
parameters equal, I achieved the results of Fig. 3 where, by setting the intraspecific coefficient equal to 1.36
(solution of the optimization problem), target species is constrained to stock values close to 50 for 20 years
before going to its maximum. It is not goal of this paper to discuss the ecological and biological meaning of
such constrained dynamics. I just want to expose here the methodological aspects of local and global networks
control.
Now let’s suppose we want to tie target species’ dynamics to pred2’s dynamics so that they are as close as
possible as time goes on, and so that they both go to their maximum (i.e. 100) at equilibrium. To this aim, I
have used equations (10) and (14) and achieved the optimized matrix of Table 2 and the optimized dynamics
of Fig. 4. As it can be seen, I was able to achieve a perfect solution (green and red lines are overlying and they
both go to 100 at equilibrium).

IAEES www.iaees.org
28 Network Biology, 2014, 4(1): 21-30

Table 2 Optimized matrix in order to tie target species’ dynamics to pred2’s ones, and so
that they both go to their maximum. The corresponding dynamics are depicted in Fig. 4.

Fig. 4 Optimized dynamics where target species and pred2 are tied together, and they both go to their maximum.

As last example, let’s suppose we want the target species to go to its maximum with a smooth dynamic (i.e.
small increments), let’s say close to 2.5 at each time step. To this aim, I used equations (10) and (16), and I
achieved the optimized matrix of Table 3 and the optimized dynamics of Fig. 5. As it can be seen, target
species increases smoothly from its initial stock to the final one as it was purposed. The target species’ initial
value was set to 43 instead of 50 by the optimization model because I stopped GAs simulations after 5 minutes,
so the methodological framework was not able to detect a perfect solution, but only something very close to it.
It’s clear from these examples that, using the proposed methodological framework, there’s no limit to the
kind of network control we can search for. It’s also clear that the same goal can be achieved via alternative and
competing optimized solutions that my framework is able to detect.

IAEES www.iaees.org
 Network Biology, 2014, 4(1): 21-30 29

Table 3 Optimized matrix that produces the corresponding dynamics depicted in Fig. 5.

Fig. 5 Optimized dynamics where target species goes smoothly to its maximum increasing by about 2.5 at each time step.

Equations from (1) top (16) are powerful enough to allow for a complete control of EBN. Not only, in fact,
we can achieve global control by attaining the desired final state, but we can also constrain local network’s
dynamics, i.e. intermediate states.
As a rule-of-thumb, I suggest that as the local control becomes more and more complex (e.g. we want to
locally and globally control more than one variable), one should use a wider control on more switches. For
instance one could start by controlling only initial stocks if the optimization goal is simple to achieve, but for
tougher goals one might enlarge the control to intra-specific and inter-specific parameters or input-output
flows as well. The result is an ecological or biological network with customized local and global dynamics.
The framework proposed here might also be applied to semi-quantitative and qualitative networks (Ferrarini,
2011b; Ferrarini, 2011c).

IAEES www.iaees.org
30 Network Biology, 2014, 4(1): 21-30

5 Conclusions
The control of EBN is a pivotal topic. Its applications are unlimited, for instance: a) neutralize damages to
ecological and biological networks, b) safeguard rare and endangered species, c) manage ecological systems at
the least possible cost, and d) counteract the impacts of climate change.
While in previous papers, I showed how to globally control EBN, here I introduced a modelling framework
for their total control, by adding local constraints to variables’ dynamics. The ratio is that sometimes
intermediate dynamics could go below or above critical ecological-biological thresholds, hence invalidating
the final global control. By locally controlling EBN, this potential drawback is fixed.

References
D’heygere T, Goethals PLM, Niels De Pauw N. 2006. Genetic algorithms for optimisation of predictive
ecosystems models based on decision trees and neural networks. Ecological Modelling, 195: 20-29
Ferrarini A. 2011a. Some thoughts on the controllability of network systems. Network Biology, 1(3-4):
186-188
Ferrarini A. 2011b. Fuzzy cognitive maps outmatch loop analysis in dynamic modeling of ecological systems.
Computational Ecology and Software, 1(1): 55-59
Ferrarini A. 2011c. Some steps forward in semi-quantitative network modelling. Network Biology, 1(1): 72-78
Ferrarini A. 2012. Network di biodiversità in alcune aree della provincia di Parma. Technical relation (in
Italian)
Ferrarini A. 2013a. Exogenous control of biological and ecological systems through evolutionary modelling.
Proceedings of the International Academy of Ecology and Environmental Sciences, 3(3): 257-265
Ferrarini A. 2013b. Controlling ecological and biological networks via evolutionary modelling. Network
Biology, 3(3): 97-105
Ferrarini A. 2013c. Computing the uncertainty associated with the control of ecological and biological systems.
Computational Ecology and Software, 3(3): 74-80
Ferrarini A. 2013d. Networks control: introducing the degree of success and feasibility. Network Biology,
3(4): 115-120
Goldberg D.E. 1989. Genetic Algorithms in Search Optimization and Machine Learning. Addison-Wesley,
Reading, USA
Holland JH. 1975. Adaptation in natural and artificial systems: an introductory analysis with applications to
biology, control and artificial intelligence. University of Michigan Press, Ann Arbor, USA
Kuo SF, Merkley GP, Liu CW. 2000. Decision support for irrigation project planning using a genetic
algorithm. Agriculture and Water Management, 45: 243–266
Liu YY, Slotine JJ, Barabasi AL. 2011. Controllability of complex networks. Nature, 473:167-173
Slotine JJ, Li W. 1991. Applied Nonlinear Control. Prentice-Hall, USA
Wall M. 1996. GAlib: A C++ Library of Genetic Algorithm Components (Version 2.4). Massachusetts
Institute of Technology, USA

IAEES www.iaees.org
 Network Biology
The Network Biology (ISSN 2220-8879) is an open access, peer-reviewed online journal that
considers scientific articles in all different areas of network biology. It is the transactions of the
International Society of Network Biology.It dedicates to the latest advances in network biology. The
goal of this journal is to keep a record of the state-of-the-art research and promote the research work
in these fast moving areas. The topics to be covered by Network Biology include, but are not limited
to:

 Theories, algorithms and programs of network analysis

 Innovations and applications of biological networks
 Ecological networks, food webs and natural equilibrium
 Co-evolution, co-extinction, biodiversity conservation
 Metabolic networks, protein-protein interaction networks, biochemical reaction networks,
gene networks, transcriptional regulatory networks, cell cycle networks, phylogenetic
networks, network motifs
 Physiological networks
 Network regulation of metabolic processes, human diseases and ecological systems
 Social networks, epidemiological networks
 System complexity, self-organized systems, emergence of biological systems, agent-based
modeling, individual-based modeling, neural network modeling, and other network-based
modeling, etc.

We are also interested in short communications that clearly address a specific issue or completely
present a new ecological network, food web, or metabolic or gene network, etc.

Authors can submit their works to the email box of this journal, networkbiology@iaees.org and(or)
wjzhang@iaees.org. All manuscripts submitted to Network Biology must be previously unpublished
and may not be considered for publication elsewhere at any time during review period of this journal.

In addition to free submissions from authors around the world, special issues are also accepted. The
organizer of a special issue can collect submissions (yielded from a research project, a research group,
etc.) on a specific topic, or submissions of a conference for publication of special issue.

Editorial Office: networkbiology@iaees.org

Publisher: International Academy of Ecology and Environmental Sciences

Address: Flat C, 23/F, Lucky Plaza, 315-321 Lockhart Road, Wanchai, Hong Kong
Tel: 00852-6555 7188
Fax: 00852-3177 9906
E-mail: office@iaees.org
Network Biology
ISSN 2220-8879
Volume 4, Number 1, 1 March 2014

Articles

Meta-analysis of cancer transcriptomes: A new approach to uncover molecularpathological

events in different cancer tissues
Sundus Iqbal, Hira Ejaz, Muhammad Sulaman Nawaz, Juan J Loor, Aisha Naeem 1-20

Local and global control of ecological and biological networks

Alessandro Ferrarini 21-30

The International Academy of Ecology and Environmental Sciences (IAEES) is a nonprofit and registered
international organization. It devotes to promote global ecology and environmental sciences and protect global
ecological environments, by publishing scientific publications, conducting research activities, launching
environmental programs, disseminating knowledge and technologies, sponsoring conferences, and providing
information and discussion spaces, etc. The responsibility for these publications rests with the International
Academy of Ecology and Environmental Sciences.

IAEES
http://www.iaees.org/