Relevance of Information in Cell Signaling Pathways using Default Logic A. Doncescu 1 , P. Siegel 2 , and T. Le 1 1 LAAS-CNRS, University of Toulouse, Toulouse, France 2 Aix Marseille Université, CNRS, LIF UMR 7279, 13288, Marseille, France Abstract— Cell Signaling Pathway Simulation is a very useful tool in the drug discovery process. These simula- tion programs can be divided into dynamic simulation and Knowledge-Based Discovery. In the first case the simulation is based on differential equations and could be considered in "real-time", meanwhile in the case of Knowledge Based Discovery Programs KBDP the consistency of the model is checked. The most efficient KBDP approach is based on first order logic (FOL). In this paper, algorithms based on Default Logic are proposed to check-out the consis- tency of the simplest representation of DNA double strand breaks. DNA double-strand breaks are among the most severe genomic lesions. This representation is concise and adequat for keeping the flow of information represented by gene expression, receptor and protein structure through the apoptosis and cell cycle. Keywords: Double Strand Breaks, DNA Damage, Default Logic, Extensions, Abduction, Consistent Pathway 1. Introduction Today the conception of artificial systems attempts to imitate the natural systems by developing new concepts of reasoning able to handle a high level of heterogeneity and uncertainty. These complex systems have a dynamic evolution in terms of structure and organization. In order to model and control these systems there is a need to observe and reconstruct their behavior by a relevant model which should make sense of large amounts of heterogeneous data gathered on various scales. System Biology is a research field, which needs an appropriate evaluation of their know- how corroborated with the available experimental data in order to represent knowledge and discover new knowledge. Therefore, System Biology could be view as a complex network constituted of protein-protein interactions, small- molecule metabolism and gene regulation. From the standpoint of Artificial Intelligence, cells are sources of information that include a large amount of intra and extra cellular signals. Disease and cancer in particular can be seen as a pathological alteration in the signaling networks of the cell. The study of signaling events appears to be the key of biological, pharmacological and medical research. For a decade signaling networks have been studied using analytical methods based on the recognition of proteins by specific antibodies. Parallel DNA chips (microarrays) are widely used to study the co-expression of candidate genes to explain the etymology of certain diseases, including cancer. The resulting data allows the modeling of gene interactions. The biological experts look for evidence of interactions between metabolites and genes. Therefore the representation by graphs is the best way to understand biological systems. This representation includes mathematical properties as con- nectivity; presence of positive and negative loops which is related to a main property of genetic regulatory networks. Biochemical reactions are very often a series of time steps instead of one elementary action. Therefore, one direction research in system biology is to capture or to describe the series of steps called pathways by metabolic engineering. All reactions that allow the transformation of one initial molecule to a final one constitute metabolic pathways. Each compound that participates in different metabolic pathways is grouped under the term metabolite. The study of gene networks poses problems well identified and studied in Artificial Intelligence over the last thirty years. Indeed, the description of network is not complete: biological experiments provide a number of protein interac- tions but certainly not all of them. On the other hand the conditions and sometimes the difficulties of the experiments involves these data are not always accurate. Some data may be very wrong and must be corrected or revised in the future. Finally the information coming from different sources and experiences can be contradictory. It is the goal of different logics, and particularly non-monotonic logics, to handle this kinds of situation. Afterwards this interaction maps should be validated by biological experiments. Of course, these experiments are time consuming and expensive, but less than an exhaustive experiment. In this paper we focus on three main problems: handling the conflicts which can occur in the gene representation, completing in-silico the gene network and the practical handling complexity of the algorithm allowing the inferences for knowledge discovery on these networks. Our approach is based on default logic allowing to handle the incomplete in- formation, and abductive reasoning to complete the missing information from the gene network. The last part is dedicated to a new language of representation, which seems to be the key to algorithm complexity handling.