From Protein Interaction Networks to Novel Therapeutic Strategies
Samira Jaeger1and Patrick Aloy1,2*
1Joint IRB-BSC program in Computational Biology, Institute for Research in Biomedicine, Barcelona, Spain
2Institucio ´ Catalana de Recerca i Estudis Avanc ¸ats (ICREA), Barcelona, Spain
Cellular mechanisms that sustain health or contribute to dis-
ease emerge mostly from the complex interplay among various
molecular entities. To understand the underlying relationships
between genotype, environment and phenotype, one has to con-
sider the intricate and nonsequential interaction patterns
formed between the different sets of cellular players. Biological
networks capture a variety of molecular interactions and thus
provide an excellent opportunity to consider physiological char-
acteristics of individual molecules within their cellular context.
In particular, the concept of network biology and its applica-
tions contributed largely to recent advances in biomedical
research. In this review, we show (i) how biological networks,
i.e., protein–protein interaction networks, facilitate the under-
standing of pathogenic mechanisms that trigger the onset and
progression of diseases and (ii) how this knowledge can be
translated into effective diagnostic and therapeutic strategies. In
particular, we focus on the impact of network pharmacological
concepts that go beyond the classical view on individual drugs
and targets aiming for combinational therapies with improved
clinical efficacy and reduced safety risks.
IUBMB Life, 64(6): 529–537, 2012
? 2012 IUBMB
protein networks; network medicine; polypharmacology;
drug repositioning; drug cocktails.
The past decade has seen a revolution in genomic technolo-
gies, and next-generation sequencing methods are delivering
fast and accurate data from genome and metagenomic projects
(1, 2). Transferring this wealth of data into biological knowl-
edge is the fundamental challenge in the postgenomic era.
Sequencing a new genome is commonly followed by a process
known as genome annotation to predict, among others, its pro-
tein coding regions and to associate biological information to
them (3). Such information is essential to understand biological
processes, cellular mechanisms, evolutionary changes, and the
onset of diseases (4, 5).
Traditional biochemical methods, such as functional assays,
knock-out experiments, or targeted mutations, have been widely
used to elucidate the role of individual gene products (6). Albeit
the function of a single gene might present a molecular descrip-
tion of cellular phenotypes, it is often not sufficient to explain
the particular processes. Thus, the question of how a single geno-
type yields distinct phenotypes remains a major challenge since
Mendel’s wrinkled peas (7) and Morgan’s white-eyed fruit flies
(8). In consequence, many fundamental biological questions
remain unanswered as traditional approaches cannot capture the
full repertoire of biochemical activities within cells mostly
because functional characteristics emerge primarily from the
complex molecular interplay between proteins, metabolites, func-
tional RNAs and genes rather than from single molecules (9).
For instance, the tumor suppressor protein p53 mediates its
natural molecular function, namely cell cycle regulation,
through several target proteins and pathways (10). Protein p53
is activated upon intra- and extracellular stimuli, such as DNA
damage, activated oncogenes, or oxidative stress. This activa-
tion induces the transcription of p53-regulated genes, e.g., Bax
or p21, through which cell cycle arrest, cellular senescence,
DNA repair, and apoptosis are initiated, depending on the phys-
iological circumstances and cell types (see Fig. 1). In turn,
mutations in p53 disrupt the complex network of stress response
pathways resulting in uncontrolled proliferation of damaged
cells and eventually to various types of cancer (11). Given its
significance as one of the most important cancer-controlling
molecules, p53 provides a primary target for cancer therapy
(12). Yet, despite intensive efforts, p53-related research did not
represent a significant breakthrough in cancer control and ther-
apy because several factors complicate the development of safe
and efficient anticancer drugs (13). Although p53 presents an
attractive drug target, from a drug design point of view it is not
Address correspondence to: Patrick Aloy, Joint IRB-BSC program in
Computational Biology, Institute for Research in Biomedicine, c/ Bal-
diri i Reixac 10-12, 08028 Barcelona, Spain. Tel: 134 934039690.
Received 13 February 2012; accepted 14 March 2012
ISSN 1521-6543 print/ISSN 1521-6551 online
IUBMBLife, 64(6): 529–537, June 2012
an ideal target as it is neither a cell surface protein nor a typical
enzyme which is chemically tractable (14). Furthermore, p53 is
known to exhibit two opposing roles in cancer progression (15).
Targeting p53 may yield strong side effects affecting also
healthy cells. Further, it is associated with premature aging and
tumor resistance. The efficacy of recent strategies, focusing for
instance on restoring p53s protective function by reactivating
mutant or activating wild type proteins are currently investi-
gated in preclinical and Phase I trials (16).
The p53 case perfectly illustrates a realistic scenario and
highlights the fact that, to elucidate the relationships between
genotype, environment and phenotype, one has to consider the
complex and nonsequential interaction patterns formed between
the different sets of cellular entities. Advanced experimental
approaches, such as DNA and protein microarrays, high-
throughput localization studies, and protein interaction mapping
methods, assess how and when these molecules interact with
each other. Several types of biological interaction networks,
such as metabolic, signaling, protein interaction, and transcrip-
tion-regulatory networks, emerge from the variety of these inter-
actions (17). Systematic studies of these networks for elucidat-
ing their basic function, structure, and dynamics have become
one of the key topics in systems biology and bioinformatics.
In this review, we focus on one of the most important types
of biological networks, i.e., protein–protein interaction net-
works, and their impact on biomedical research. We first present
network-based approaches for unveiling disease mechanisms, a
prerequisite for conducting effective therapeutic treatment. Fur-
ther, we outline how the detailed knowledge on pathological
pathways may be directly translated into therapeutic strategies
that treat symptoms and causes of common complex diseases.
In this part, we focus particularly on current advances in net-
PROTEIN INTERACTIONS AND NETWORKS
Protein interaction networks present gene products that
physically interact with each other to accomplish particular cel-
lular functions, such as metabolism, cell cycle control, and sig-
nal transduction (18). A number of experimental techniques
have been developed for studying these interactions and their
characteristics in small- and in large-scale, see (19, 20) for a
review. Traditionally, protein interactions have been identified
by classical genetic, biochemical or biophysical methods, such
as fluorescence resonance energy transfer or X-ray crystallogra-
phy. Such small-scale studies focus on selected proteins for
generating specific interaction maps (21–23). However, the
increasing availability of fully sequenced genomes and the
speed at which proteins are discovered facilitated the progress
of technologies that screen large sets of candidates systemati-
cally. Two widely established large-scale methodologies are the
yeast two-hybrid (Y2H) system (24) and tandem affinity purifi-
cation coupled to mass spectrometry (TAP-MS) (25). Both tech-
niques have been used frequently for large-scale experiments in
different model organisms, including yeast, fly, worm, and
Despite being incomplete (26) and error-prone (27), system-
atic studies of protein interaction networks have been proven to
be particularly important for deciphering the relationships
between network structure and function (28), discovering novel
protein function (29), identifying functionally coherent modules
(30, 31), and conserved molecular interaction patterns (32, 33).
In addition, interaction networks have become essential and
powerful tools for associating proteins with distinct phenotypes
and diseases (34, 35), as well as for studying pharmacological
drug-target relationships (36, 37).
Although we focus in this review on classical physical pro-
tein interaction data, there are several other techniques for elu-
cidating molecular relationships, e.g., synthetic genetic arrays
(38) and RNA interference (39). Genetic but also functional
interaction data obtained from such methods can be used similar
as or integrated with protein interaction data (40, 41).
Figure 1. The p53 network. p53 is one of the central compo-
nents of the complex network of stress response pathways
(Adapted from Ref 10). The activation of the network upon
DNA damage, stress or activated oncogenes initiates the modifi-
cation of p53 and its negative regulator MDM2. Activated p53,
in turn, induces the expression of the target genes, such as p21,
Bax, or Fas, to mediate its various functions including cell cycle
arrest, DNA repair, apoptosis, and senescence. Potential com-
pounds for restoring p53s function in cancerous cells are shown
in orange (preclinical trial) and yellow (Phase I trial).
530 JAEGER AND ALOY
72. von Eichborn, J., Murgueitio, M. S., Dunkel, M., Koerner, S., Bourne,
P. E., et al. (2011) PROMISCUOUS: a database for network-based
drug-repositioning. Nucleic Acids Res. 39, D1060–D1066.
73. Campillos, M., Kuhn, M., Gavin, A. C., Jensen, L. J., and
(2008) Drug target identification using side-effect similarity. Science
74. Kuhn, M., Campillos, J. M., Bork, P., Jensen, L. J., Gavin, A. C., et al.
(2008) Use of aprepitant and derivatives thereof for the treatment of
cancer. Patent WO 2009/124756 A1; CT/EP2009/002621.
75. Duran-Frigola, M. and Aloy, P. (2012) Recycling side-effects into clin-
ical markers for drug repositioning. Genome Med. 4, 3.
76. Sundar, S., Jha, T. K., Thakur, C. P., Engel, J., Sindermann, H., et al.
(2002) Oral miltefosine for Indian visceral leishmaniasis. N. Engl. J.
Med. 347, 1739–1746.
77. Rajkumar, S. V. (2004) Thalidomide: tragic past and promising future.
Mayo Clin. Proc. 79, 899–903.
78. Keith, C. T., Borisy, A. A., and Stockwell, B. R. (2005) Multicompo-
nent therapeutics for networked systems. Nat. Rev. Drug Discov. 4, 71–
79. Jia, J., Zhu, F., Ma, X., Cao, Z., Li, Y., et al. (2009) Mechanisms of
drug combinations: interaction and network perspectives. Nat. Rev.
Drug Discov. 8, 111–128.
80. Tsang, R. Y. and Finn, R. S. (2012) Beyond trastuzumab: novel thera-
peutic strategies in HER2-positive metastatic breast cancer. Br. J. Can-
cer 106, 6–13.
81. Nahta, R. and Esteva, F. J. (2006) HER2 therapy: molecular mecha-
nisms of trastuzumab resistance. Breast Cancer Res. 8, 215.
82. Gupta, G. P., Nguyen, D. X., Chiang, A. C., Bos, P. D., Kim, J. Y.,
et al. (2007) Mediators of vascular remodelling co-opted for sequential
steps in lung metastasis. Nature 446, 765–770.
83. Ainsworth, C. (2011) Networking for new drugs. Nat Med. 17, 1166–
537 FROM INTERACTION NETWORKS TO THERAPEUTIC STRATEGIES