Identification of critical regulatory genes in cancer signaling network using controllability analysis

https://doi.org/10.1016/j.physa.2017.01.059Get rights and content

Highlights

  • Controllability analysis of cancer regulatory mechanisms to identify ‘driver genes’.

  • Indispensable backbone driver genes critical for steering the state of the network.

  • Critical genes are associated to cancer and are targets of antineoplastic drugs.

  • Driver genes as mediators for driving cancer phenotype into a healthy state.

  • Insights into cancer mechanisms and means for identification of drug targets.

Abstract

Cancer is characterized by a complex web of regulatory mechanisms which makes it difficult to identify features that are central to its control. Molecular integrative models of cancer, generated with the help of data from experimental assays, facilitate use of control theory to probe for ways of controlling the state of such a complex dynamic network. We modeled the human cancer signaling network as a directed graph and analyzed it for its controllability, identification of driver nodes and their characterization. We identified the driver nodes using the maximum matching algorithm and classified them as backbone, peripheral and ordinary based on their role in regulatory interactions and control of the network. We found that the backbone driver nodes were key to driving the regulatory network into cancer phenotype (via mutations) as well as for steering into healthy phenotype (as drug targets). This implies that while backbone genes could lead to cancer by virtue of mutations, they are also therapeutic targets of cancer. Further, based on their impact on the size of the set of driver nodes, genes were characterized as indispensable, dispensable and neutral. Indispensable nodes within backbone of the network emerged as central to regulatory mechanisms of control of cancer. In addition to probing the cancer signaling network from the perspective of control, our findings suggest that indispensable backbone driver nodes could be potentially leveraged as therapeutic targets. This study also illustrates the application of structural controllability for studying the mechanisms underlying the regulation of complex diseases.

Introduction

Cancer is a complex disease which is characterized by subtle interplay of regulatory mechanisms underlying its phenotype  [1], [2]. Dysregulation of multiple pathways governing fundamental cell processes (such as death, proliferation, migration and differentiation) is known to be a key cause for emergence of cancer. The crosstalk between signaling pathways reflecting salient aspects of disease have been used to model this pathology  [3], [4]. The focus behind building such integrative models has been to create a meaningful molecular picture of cancer so as to find ways for controlling the disease  [5], [6], [7].

Systems that exhibit complex phenomena owing to interconnected mechanisms underlying its architecture can be studied using graph theoretical paradigm  [8], [9], [10]. Study of such networked systems of social, technological and biological origin has added to the understanding of their structure, function and evolution. Availability of rich data mapping biological processes in the postgenomic era has facilitated creation of molecular interaction catalogues (protein interactomes, gene regulatory networks, metabolic pathways and co-expression networks), and better understanding of cellular functions [11], [12], [13]. Using these graph theoretical approaches, various studies have attempted to identify regulatory mechanisms which are central to the disease  [14], [15], [16], [17]. Rapid advances in network biology have provided a new conceptual framework revolutionizing the view of biology and disease pathology.

From control systems perspective, cellular processes can be viewed as an intricately controlled orchestra of regulatory mechanisms that lend the cell its functional repertoire. Diseases, therefore, can be seen as the result of errors in cellular information processing. Beyond systems modeling of diseases, the focus has also been on finding ways of controlling the disease through therapeutic interventions. Recent excitement in controllability of networks and identification of driver nodes as agents for steering the state of the network has added much needed impetus in this direction  [18], [19], [20], [21], [22]. Control systems theory proposes that the state of complex networks can be controlled with the help of a set of ‘driver nodes’. Driver nodes may then be fed with external inputs to steer the state of the network.

Analysis of systems biological models of diseases has provided crucial insights into their mechanisms and potential drug targets  [4], [23]. The understanding garnered through such studies has often taken a static perspective of disease interactomes, ignoring dynamical aspects. Study of controllability of diseases and search for driver nodes as potential therapeutic targets provides a new dimension. Such integration of control theory with disease interactomes could pave way for better strategies to assist drug discovery process as well as improve our understanding of the disease  [24], [25], [26].

Cancer systems biology has proved to be helpful in implementing new therapeutic strategies  [5], [27]. While most studies  [23], [28], [29] focus on identification of epigenetic changes and mutations in genes, and target them as means of controlling the disease, control theory offers a framework for arriving at driver genes that could be used for steering the state of the regulatory network. We propose that identification of specific molecules as drivers of regulatory dynamics could be a promising step towards targeted cancer therapies.

Here, we modeled the human cancer signaling network as a directed graph and probed for its critical regulators using structural controllability. We implemented the maximum matching algorithm to identify driver nodes. The driver nodes were divided into backbone, peripheral and ordinary based on their role in regulatory interactions and control of the network. Based on node deletion studies that enumerate impact of a node on ease of network control, indispensable nodes were identified. These indispensable backbone driver nodes were found to be critical for driving the regulatory network into cancer phenotype (via mutations) as well as for steering into healthy phenotype (as drug targets). Thus they emerged as central to control, both as causal elements by virtue of mutations and also as therapeutic agents in the form of cancer drug targets. This study illustrates an application of control theory for investigation of regulatory mechanisms underlying complex diseases.

Section snippets

Human cancer signaling network

We created the Human Cancer Signaling Network (HCSN) starting from the data of human signaling network  [30] comprising of 1634 genes and 5089 regulatory relations by integrating genetically and epigenetically altered cancer associated genes and signaling pathways for cancer. Thus HCSN embeds molecular correlates associated to cancer. The nodes in HCSN are signaling molecules (such as genes, proteins and other small molecules) and links represent effector actions such as activation or

Topological characterization of HCSN driver nodes

HCSN is a molecular interactome of signaling mechanisms of cancer (Fig. 1). Driver nodes, that facilitate control of the network state, could provide means of controlling the dynamical state of cancer associated signaling network. Using maximum matching algorithm, we identified driver nodes and classified them into PDNs, ODNs and BDNs, depending on their presence in MDNS across multiple sample sets. A total of 584 nodes (out of 1232) were driver nodes with 29.38% PDNs, 31.01% ODNs and 39.61%

Discussion and conclusions

Our study integrates the systems biological approach to cancer regulatory mechanisms with control theory to identify biological implications of ‘driver nodes’. We propose the notion of regulatory network as an underlying molecular framework that is subject to control through indispensable backbone driver nodes. These nodes could either steer the network from healthy state to disorder by means of mutations, or could be leveraged as drug targets for driving the network into healthy state. These

Acknowledgments

VR acknowledges the Inspire Fellowship from Department of Science and Technology (DST/INSPIRE/IF120809). VR, VS and GB acknowledge DA-IICT for providing computational facility. GB also acknowledges support from IIIT-Delhi.

References (33)

  • M. Kanehisa et al.

    KEGG for integration and interpretation of large-scale molecular data sets

    Nucleic Acids Res.

    (2011)
  • S.A. Forbes et al.

    Cosmic: exploring the world’s knowledge of somatic mutations in human cancer

    Nucleic Acids Res.

    (2015)
  • K. Tomczak et al.

    The cancer genome atlas (tcga): an immeasurable source of knowledge

    Contemp. Oncol.

    (2015)
  • A.-L. Barabasi et al.

    Network biology: understanding the cell’s functional organization

    Nature Rev. Genet.

    (2004)
  • V. Randhawa et al.

    Identification of src as a potent drug target for asthma, using an integrative approach of protein interactome analysis and in silico drug discovery

    OMICS: J. Integr. Biol.

    (2012)
  • S. Vashisht et al.

    An approach for the identification of targets specific to bone metastasis using cancer genes interactome and gene ontology analysis

    PLoS One

    (2012)
  • Cited by (0)

    View full text