Modelling & Scientific Computing Group
The Modelling & Scientific Computing Group focuses mainly on computational models/ methods and analysis, and scientific applications of computing. Computational science problems range from condensed matter and statistical physics, through socioeconomic systems, to the biomedical sciences. The group has expertise in mathematical modelling and statistical analysis, employs a range of computational modelling methods, including top-down and bottom-up (CA, MC, agent-based), and utilises parallel and distributed computing for analysis and simulation.
Recent efforts included participation in the DREAM5* consortium challenge for robust gene network inference from high throughput data. The so-called 'Wisdom of Crowds' approach exploits the collective knowledge base to address long standing challenges, such as that of finding and modelling gene regulatory networks. The 5th annual systems biology computational challenge enabled characterisation and assessment of a range of inferential methods (including the evolutionary algorithm approach contributed by the DCU group), on benchmark datasets, with community-based 'concensus networks' found to be robust across data sources and type, (Published Nature Methods 2012 - http://modsci.computing.dcu.ie/publications.shtml ) The wider relevance of such challenges is clear, with tractability of large-scale datasets a current focus for many disciplines and integrative data analysis for predictive modelling offering major advantages.
*Dialogue on Reverse Engineering Assessment and Methods
There are currently a number of research themes in the Modelling &Scientific Group: http://modsci.computing.dcu.ie
Biocomputation (with research on Bio-systems modelling, Bioinformatics/Bio-data analytics, Biometrics, Models of disease, Bio-diversity and AI - for bio and artificial systems and Pattern Recognition).
Financial and Socioeconomic Modelling (including projects in Econophysics, statistical and computational models of Finance, system flows (social nets, traffic) and multivariate techniques in Finance). Business Informatics/analytics affiliation: http://www.computing.dcu.ie/big/
Sub-group in Environmental Modelling (including expertise in Wind and Wave Energy and Pollution Modelling, Pedestrian flows and urban impact, Green cityscapes - heterogeneous non-motorised/motorised traffic mix).
Affiliated: 3D Vision Group (Computer Vision, Human Behaviour Recognition using Mutliple Cameras, Autonomous Vehicle Navigation, 3D Object Recognition, Shape Recognition, Digital Learning using the Kinect). http://3dvg.computing.dcu.ie
Assisted Living/Social Behaviour: in liaison with CLARITY centre. Current project on modelling & analysis of Lifelogging data: signal processing/bio-inspired pattern analysis.
Example Research Projects
- Biocomputation - Drug Dissolution Modelling: the origins of this project lie in an EU FP4-funded project in collaboration with TCD and Elan Corporation to model drug dissolution in vitro. Funding continued under the Biocomputing Strand of the National Institute for Cellular Biotechnology under PRTLI 3, to demonstrate the application of probabilistic and semi-analytical methods to in vitro drug dissolution for a wider variety of drug delivery devices and conditions. Further funding has been secured to extend the model to the design of Therapeutic Implants. DCU’s role is to simulate dissolution and cellular ingress in the implant, with resulting changes in implant mechanical properties, offering possibilities for micro level targeted treatment. With the introduction of Bayesian Inference, direct and inverse Monte Carlo and other probabilistic numerical methods, we believe the results show enormous potential for the simulation of in vivo targeted drug delivery simulation, a ‘holy grail’ of drug development and something of potentially huge benefit to the Pharmaceutical Industry.
Biocomputation - Epigenetics Modelling: Funded under FP6 and Complexity Science continuation (FP7), as well as support from SFI and IRCSET, the work on computational epigenetics addresses modelling of changes to phenotype, due to modifications in gene expression without alteration to the DNA sequence. The complex interdependent mechanisms involved define the epigenome layer, which influences function of the genome. Links to epigenetic "signatures", such as DNA methylation, histone modification and changes in chromatin, have been established in cancer, in autoimmune and neuropsychiatric disorders, in response to stress and also in the ageing process. Epigenetic signatures often match to differential or abnormal gene expression profiles, with epigenetically-triggered silencing, or over-expression of genes. Investigative strands, as below, have each led to publication, most recently on implications for cancer development under different DNA methylation inhibition profiles.
- An epigenetic signature level micro-model.
- A representation of infection-induced aberrant DNA methylation in gastric cells,(developed in collaboration with the National Cancer Center (Tokyo, Japan), and used as a proof of concept).
- StatEpigen database, a novel knowledge management system for genetic and epigenetic molecular determinants of cancer.
- A network-based model for micro-molecular events observed at different stages of colon cancer, with a focus on the gene relationships and tumour pathways.
- Social Systems - Environmental/Urban Networks: Variously funded by IRC awards and extension of an EU Complexity Science project, the work on urban movement, (traffic and pedestrian flow and transport modalities), has seen development of models for heterogeneous traffic, both motorised and non-motorised (with bicycle stream for ‘green city’ impact) and latterly of ‘cityscaping’ for pedestrian flows. Pedestrian movement must be adaptable to fast decision changes, with individual and crowd behaviour, as well as street-space perception, important for journey dynamics, and influential in provision and use of facilities, evacuation and flow management. Simulation methods, such as stochastic cellular automata, the agent-based paradigm, amongst others, can be used as modelling tools to explore random choice, interaction of individuals and the environment, vehicle manoeuvres and flow dynamics. This can enhance the network picture of static GIS, while databases, such as OSM, facilitate sourcing of complex spatial environment data.