by Irina-Afrodita Roznovăţ and Heather J. Ruskin
The aim of our current work is to investigate the interdependencies of genetic and epigenetic mechanisms leading to aberrations in cancer initiation and progression. The objectives are to develop a computational model for colon cancer dynamics, linking microscopic effects to macroscopic outcomes, and to analyse the impact of different risk factors on malignant tumour development.
{jcomments on}Conducted at the Centre for Scientific Computing & Complex Systems Modelling, (SCI-SYM), School of Computing, Dublin City University, this research is being funded by the CIESCI ERA-Net Complexity Project, (EC/Irish Research Council and as an extended Embark project (IRC)). The project “Complexity of Interdependent Epigenetic Signals in Cancer Initiation” is a collaboration between three countries and laboratories: SCI-SYM, Dublin City University, Dublin, Ireland; the Bellvitge Institute for Biomedical Research, (IDIBELL), Barcelona, Spain, and the Bioinformatics, Algorithmics, and Data Mining Research Group (BIIT), Institute of Computer Science and Estonian Biocenter, University of Tartu, Estonia.
Defined as micro-molecular interactions that influence gene expression without altering the DNA sequence, epigenetic events have been detected (i) in the earliest stages of neoplastic disease initiation, (ii) in the ageing process and also (iii) in response to cellular stress. Understanding these phenomena, which are considered markers for tumour initiation, may increase the success of cancer therapy in affected patients. Given time and cost implications for laboratory experimentation and human genome studies, a range of computational models has been developed over recent decades, to help scientists and clinicians to better understand the impact of abnormal micromolecular modifications, observed to occur in neoplastic diseases. Different computational methods such as artificial neural networks, support vector machines and hidden Markov models, inter alia, have been applied to analyse and predict the patterns of genetic and epigenetic signals in malignant systems. Additionally, databases that focus on various aspects of tumour pathways are increasingly being developed and populated with specific data. Moreover, these sources capture information on epigenetic “signatures”, such as DNA methylation, histone modification and changes in chromatin, together with genetic mutations of cancer-related genes.
The project aims to develop a multi-layered model to study colorectal cancer dynamics, incorporating a Bayesian network approach for estimation of incomplete information [1]. The model structure is given by the interdependencies between three main layers: micromolecular signals, gene relationships and cancer stage transitions. As a basis, the gene framework integrates empirical data on conditional relationships, between genetic and epigenetic events found in colon cancer development. The gene network is then allowed to evolve over time. After a number of iterations, which reflect cellular time scales, methylation level (for example) is checked for the entire gene network and provides a major input to the decision on cancer stage attained. A schematic outline of key influences and flows in the simple model is given in Figure 1.
The colon cancer model utilises an object-oriented framework (written in C++). Initially developed and tested in a single-threaded setting, preliminary results clearly indicated the need for speed-up, as well as inclusion of more complex computational flows. Parallelisation of the algorithm is thus the current focus, using the existing framework, (as outlined) and the SCI-SYM in-house computing cluster. Up-scaling to national facilities at the Irish Centre for High-End Computing, (ICHEC), is also anticipated.
Realistically, predictive model development requires integration of additional risk factors, (such as ageing, gender, heredity), and some way of quantifying the influence of these on disease progression. In consequence, the gene framework is being expanded to include and assess environmental and lifestyle impact, as well as extended mutation information. Moreover, identification of a commonly mutated/methylated gene base, offers the potential for the colorectal cancer model to be extended to other cancer types, such as stomach, lung, or liver.
SCI-SYM ‘s contribution to the CIESCI project complements its other work on computational modelling for molecular events in malignant diseases and related bioinformatics analyses. An agent-based model has been developed to determine the risk of gastric cancer, from aberrant DNA methylation levels induced in cells, following infection with Helicobacter pylori, (providing further insight on epigenetic events in abnormal conditions). Additionally, the StatEpigen database, [2], containing manually annotated and curated data from primary and secondary sources, has been built to provide specific information on genetic and epigenetic event determinants of gene relationships at different cancer stages.
Acknowledgements
IR is grateful for financial support from the CIESCI ERA-Net Complexity Project, (EC/IRCSET) for 2 years and final year top-up under Irish Research Council (IRC).
Links:
http://sci-sym.computing.dcu.ie/
http://computing.dcu.ie/
http://www.computing.dcu.ie/~msc/
Reference:
[1] I.A. Roznovat and H.J. Ruskin: “A Computational Model for Genetic and Epigenetic Signals in Colon Cancer,”, in proc. IEEE BIBMW, Philadelphia, USA, pp. 188 – 195, 2012
[2] A. Barat and H.J. Ruskin: “A Manually Curated Novel Knowledge Management System for Genetic and Epigenetic Molecular Determinants of Colon Cancer”, Open Colorectal Canc. J., 3, pp. 36 – 46, 2010.
Please contact:
Irina-Afrodita Roznovăț
Dublin City University, Ireland
Tel: +353 1 700 6747
E-mail:
Heather J. Ruskin
Dublin City University, Ireland
E-mail:
