EPSRC Network on Evolvability in Biology & Software Systems

Evolvability, Genetics & Development in Natural and Constructed Systems: Abstracts of the EPSRC Evolvability Network Symposium

Tewin Bury Farm Hotel, Hertfordshire, England, UK
26-28 August 2003

University of Hertfordshire Computer Science Technical Report 389
C. L. Nehaniv, P. J. Bentley & S. Kumar (Editors)

Feasibility of Automating the Process

Feasibility of Automating the Process of Genetic Regulatory Network Inference


Biocomputation Research Laboratory
Science and Technology Research Centre
University of Hertforshire
College Lane
Hatfield, Hertfordshire AL10 9AB

I.Abnizova@herts.ac.uk, C.Battail@herts.ac.uk

In bilaterian animals such as ourselves, developmental process are controlled by large and complex genetic regulatory networks. Regulatory interactions between genes are primarily mediated through cis-regulatory elements. The architecture of the network is defined by the functional linkages from given regulatory genes to the cis-regulatory elements (i.e., inputs) of other genes. This interconnection determines the deployment of sets of genes in developmental space and time.

A logic map of intergenic regulatory interactions has been developed by the Davidson's lab (Division of Biology, Caltech, USA), in collaboration with the Biocomputation group (University of Hertfordshire, Hatfield, UK), for the network that controls endomesoderm specification in sea urchin embryos. This model comprises almost 50 interacting genes at present and predicts the potential functional target sites of the cis-regulatory regions. The network architecture was constructed on an ad-hoc basis. Most of the functional linkages were discovered from a large-scale perturbation analysis in which the expression of many regulatory genes and several signalling pathways were altered experimentally. The effects on many other genes were then measured using quantitative polymerase chain reaction (QPCR). Temporal and spatial information of gene expression determined by QPCR and whole mount in situ experiments, experimental cis-regulatory data, results of rescue experiments and knowledge from experimental embryology were also combined.

Our overall objective is to develop a computational framework which combines constraints arising from these multiple information sources to automatically infer the structure and parameters (linkage probabilities) of the gene regulatory network controlling the endomesoderm specification in sea urchin embryo.

An intermediate stage is evaluated first in order to analyze the feasibility of the global approach. We are developing a computational framework to learn linkage probabilities of a regulatory network model knowing its structure and we apply it on experimental data used to identify the endomesoderm specification network. Statistical dependencies are determined between each pair of genes from each sources of experimental data (temporal gene expression profiles, spatial expression data and, perturbation analysis data) using a combination of normalization methods, dissimilarity measures, clustering algorithms and, validity criterions. Correlation networks between genes, generated by this approach, present strong structural similarities with the endomesoderm Boolean model. Statistical methods were therefore able to extract relevant gene dependencies from raw data.