Reasoning and hypothesizing about signaling networks
Chitta R. Baral, Nam Tran · 2006
A living cell constantly receives and responds to signals from its environment. Signals come in the form of light, heat, or biochemical molecules. Signals initiate networks of biochemical reactions in the cell, which result in cell growth, cell movement, cell survival or programmed cell death. Malfunctions of these signaling networks usually cause diseases. For example, a signal may not be turned off after it has triggered cell division, thus resulting in excessive growth of cells and causing cancer. Consequently, effective therapeutic strategies may be based on correction or adjustment of broken-down signaling networks. The goal of this dissertation is to contribute toward computer-aided modeling and inference of signaling networks. Most of the existing approaches are based on quantitative models such as differential equations. The analysis of a model is usually done via simulation and perturbation. Such an analysis can easily answer prediction queries (about impact of events on the cell). However, it is computationally expensive to answer questions about explaining a particular observation, or about planning to alter the cell behavior. Moreover, quantitative information (e.g., kinetic data) is hard to obtain, thus rendering the quantitative models inexact. Several qualitative approaches have been proposed, but they have not yet addressed explanation and planning queries. Besides, they do not deal with the incompleteness of information about signaling networks, such as missing pathways or uncertainty of observations. In this dissertation, a knowledge-based approach is proposed for representing and reasoning about signaling networks. A non-monotonic knowledge representation language is presented that can handle incomplete or partial information about signaling networks. The language and a prototype of knowledge have been implemented using a state-of-the-art AnsProlog reasoning engine, which can compute prediction, explanation and planning queries. The proposed approach also allows for easy updating of the knowledge base when new knowledge becomes available.