MODELING MOLECULAR PATHWAYS BASED ON GENE EXPRESSION AND SOCIAL NETWORK ANALYSES: AN EXAMPLE FROM ARABIDOPSIS THALIANA

Although the functional complexity of biological systems still requires suitable platforms and dedicated analyses to unravel details on the involved components and on mechanisms for their regulation and control, the design of synthetic bio-models able to reproduce useful functionalities is now becoming concrete. To this end, however, exploiting resources from complex organisms is still a challenge. As an example, the whole genome sequence of Arabidopsis thaliana in 2000 revealed that, despite the suitability of this organism as a model for plant biology, its small diploid genome showed an unexpected complexity in terms of intragenome duplications. This may complicate the identification of key genes involved in a functional network among the many variants the whole genome encodes. Moreover, although the role of Arabidopsis as a reference in plant biology and its manifold related resources, the lack of a fulfilled gene annotation and of exhaustive technological platforms still weakens the identification of all the partners in a molecular pathway. We considered the gene expression values from microarray results on Arabidopsis tissues collected in the AtGenExpress Project Atlas at the NASC website (http://affymetrix.arabidopsis.info/narrays/experimentbrowse.pl), filtering out the experiments in physiological conditions. Pearson correlation tests permitted to define global gene correlations highlighting that i) about 83% of the genes have at least one correlation with another gene and that ii) inverse correlations are less frequent than direct ones. In order to identify among correlated genes those depicting specific functional networks we applied a novel approach. This latter, based on social network analyses, investigates on relationships among genes in terms of the network theory. Gene networks can be filtered out on the base of the number of direct correlations shared among genes that belong to a cluster of correlated ones. Suitable thresholds in the applied methodology, together with gene family analyses and annotation clues, highlighted upscale gene associations, which provide further insights to improve the accuracy of network descriptions. Our preliminary results underline the strong need of these approaches to identify the effective components of a functional pathway