From Genome to Life:

THIEFFRY Denis

University of Aix-Marseille II (ESIL-GBMA), 13, Av. de Luminy, Marseille 13288, France

title: Computational integration, analysis, and simulation of genetic regulatory networks.

As molecular data about gene structure, expression and regulation are accumulating at an ever increasing pace, it becomes urgent to develop appropriate methods to check our understanding of the global dynamics of gene expression. Indeed, specific patterns of gene expression result from complex regulatory networks, involving many regulations at various levels. Such regulatory networks can be represented in terms of graphs of interactions and thus analysed with the help of graph-theoretical tools and concepts. However, we do need additional formal tools to represent, analyse, and simulate the dynamical properties of the corresponding networks. As most data presently available are of qualitative nature, we use a "logical" formalism to distinguish discrete levels of gene expression (using logical variables and functions which can take two or more values) and to evaluate gene interactions (using logical parameters). The flexibility and power of our logical formalism is illustrated by the modelling and the analysis of the gap gene network involved in Drosophila melanogaster segmentation. The gap genes are expressed in defined domains along the anterior-posterior axis of the embryo, as a response to asymmetric maternal information in the oocyte. Though many of the individual interactions among maternal and gap genes are reasonably well understood, we still lack a thorough understanding of the dynamic behaviour of the system as a whole. Developed in collaboration with Lucas Sánchez (CIB, Spain), our model analysis leads to the delineation of: (1) the minimal number of distinct, qualitative, functional levels associated with each of the key regulatory factors (the three maternal Bcd, Hb and Cad products, and the four gap Gt, Hb, Kr and Kni products); (2) the most crucial interactions and regulatory circuits of the earliest stages of the segmentation process; (3) the ordering of different regulatory interactions governed by each of these products according to corresponding concentration scales; and (4) the role of gap-gene cross-interactions in the transformation of graded maternal information into discrete gap-gene expression domains. Not only does the proposed model allow a qualitative reproduction of the patterns of gene expression experimentally characterised, but it allows also the qualitative simulation and prediction of the phenotypes of single and multiple loss-of-function mutations, of cis-regulatory mutations, as well as of ectopic gap gene expression.