Formal Modelling and Analysis of the role of Hexosamine Biosynthetic Pathway in Cancer: Exploiting Parallelism in Qualitative Biological Regulatory Networks

Abstract

Background: Cancer is one of the leading causes of death worldwide. With advancements in high throughput technologies and availability of data on gene regulations, knowledge about progression of cancer has improved and it is now viewed as a complex multifaceted systems level disease. Despite heterogeneity of the malignancies, key functions in development of cancer are common. The alteration of glucose metabolism is considered as important hallmark of cancer and an essential factor towards cell growth and invasion. Increased flux of glucose through the Hexosamine Biosynthetic Pathway (HBP) drives increased cellular O-GlcNacylation and contributes to cancer progression. However, the role of HBP in activation of key oncogenes and progression of cancer is poorly characterized. In this study, a systems-biology approach based on qualitative modelling framework (proposed by René Thomas) is used to investigate the role of HBP in activation of oncogenes that lead to cancer progression. In qualitative modelling approach, dynamic behavior of the system under investigation is determined by model parameters which are not known in advance. The parameter estimation required for qualitative modelling is computationally intensive task and takes lot of processing time. By using parallel computing, we address computationally challenging aspects of qualitative modelling which involve parameters estimation and identification of important trajectories in the model. Methodology: The methodology used to investigate the role of HBP in progression of cancer is based on qualitative modelling, model-checking, network 1 Abstract analysis using betweenness centrality and petri net modelling approach. First, in order to construct a qualitative model, key regulatory entities from the literature are incorporated in the model. Second, model parameters are computed from observations using the model-checking technology. The total number of model parameters increases exponentially with increase in number of entities. In order to cope with the complexity of parameters estimation for qualitative modelling, we use a Java based software MPJ-Express for parallelization of sequential implementation of SMBioNet software. The parallel approach divides the parameter space into different regions and each region is explored concurrently on multicore and cluster platforms. Third, from the computed set of model parameters, a dynamic qualitative model is constructed by using GinSim software. The dynamic model cannot be analyzed manually because of large number of trajectories. Hence, centrality based network analysis is carried out by using Cytoscape software for identification of important trajectories. These trajectories are further analyzed by investigating step-by-step alterations in gene expressions that lead to activation of key oncogenes and development of cancer. Fourth, hybrid modelling is carried out to compute delay constraints using HyTech model-checker. These delay constraints highlight logical relationships between synthesis and degradation rates of important genes in the model. Finally, a stochastic petri net model is developed using Snoopy software for validation of delay constraints. Results: The experimental results indicate O-linked N-acetylglucosamine transferase (OGT) as a key regulator that promotes oncogenesis in a feedback mechanism through the stabilization of C-Myc. The absence of p53-Mdm2 oscillation 2 Abstract is identified as another important contributor towards progression of cancer. Silencing of OGT and C-Myc loop reduces the glycolytic flux, while restoration of P53-Mdm2 oscillations leads to recovery and restoration of homeostasis. Together, our findings suggest potential targets that may provide a mechanismbased therapeutic approach for regulation of hyper-O-GlcNacylation in human cancer. The parallel approach presented in the study reduces processing time for parameter estimation for our qualitative model of the Hexosamine Biosynthetic Pathway and achieves almost linear speed-up on multicore and cluster platforms . The Parallel-SMBioNet implementation for logical parameters estimation is provided at http://systemsbiology.tools. Conclusion: We use a formal modelling approach to study the function of the Hexosamine Biosynthetic Pathway, which triggers hyper O-GlcNAcylation. Within the p53-Mdm2 circuit, we compute important delay constraints involving synthesis rates in order to restore homeostasis. We analyze different simulation trajectories, which showed that enhanced expression of O-GlcNAc-transferase (OGT) consistently upregulates NF-κB, PI3K and FoxM1. Moreover, persistent activation of OGT through c-Myc drives the system to a deadlock state from where recovery is not possible. These findings suggest that OGT is acting as a critical mediator of various oncogenic and tumor suppressor proteins implicated in tumor growth and development. We acknowledge that our findings are derived from a qualitative approach and could be dependent on cellular dynamics and environment. However, these discoveries form the foundation and direction of future translational research studies to design a quantitative model with additional 3 Abstract tools and experimental verification for the development of molecular therapeutics. Taken together, mechanism-based therapies that are designed to target hyper O-GlcNAcylation and OGT may hold clinical benefits in the treatment of cancer.