ECAI 2004 Conference Paper

[full paper]

Towards a Logical Analysis of Biochemical Pathways

Patrick Doherty, Steve Kertes, Martin Magnusson, Andrzej Szalas

Biochemical pathways or networks are generic representations used to model many different types of complex functional and physical interactions in biological systems. For example, metabolism can be viewed and modeled in terms of complex networks of chemical reactions catalyzed by enzymes and consisting of reactive chains of substrates and products. Often these models are incomplete. For example, reactions may be missing and only some products are observed. In such cases, one would like to reason about incomplete network representations and propose candidate hypotheses, which when represented as additional reactions, substrates, products, or constraints on such, would complete the network and provide causal explanations for the existing observations. In this paper, we provide a logical model of biochemical pathways and show how hypothesis generation may be used to provide additional information about incomplete pathways. Hypothesis generation is achieved using weakest and strongest necessary conditions for restricted fragments of 1st-order theories which represent these incomplete biochemical pathways and explain observations about the functional and physical interactions being modeled. Quantifier elimination techniques are used to automatically generate these hypotheses. Part of the modeling process includes the use of approximate databases where inferences can be made about the resultant hypotheses.The techniques are demonstrated using metabolism and molecular synthesis examples.

Keywords: Knowledge Representation, Bioinformatics, Abduction

Citation: Patrick Doherty, Steve Kertes, Martin Magnusson, Andrzej Szalas: Towards a Logical Analysis of Biochemical Pathways. In R.López de Mántaras and L.Saitta (eds.): ECAI2004, Proceedings of the 16th European Conference on Artificial Intelligence, IOS Press, Amsterdam, 2004, pp.997-998.