Knowledge Representation, Reasoning, and Problem Solving in a Cellular Domain

Arizona State University, Scottsdale AZ

Investigators

Abstract

The aim of the proposed research is to develop new and adapt existing knowledge representation and reasoning schemes to represent and perform various kinds of reasoning and problem solving tasks, such as prediction, explanation, diagnosis and planning about the activities inside a cell. This project aims to focus on happenings and mechanisms such as signal transduction, protein-protein interactions and other biochemical reactions. The project's preliminary results suggest that the knowledge representation, reasoning, and problem solving approaches of Artificial Intelligence are well suited for this task. In a cell when a ligand binds to an appropriate receptor it activates the receptor which in turn triggers a series of events. In the terminology of planning and reasoning about actions, the binding of a ligand to the receptor can be thought of as an action; the conditions under which a ligand can bind to a receptor, as the executability conditions of an action in a planning domain; Similarly, the direct effects of the ligand binding to a receptor can be expressed using planning terminologies. Planning is just one of the problem solving and reasoning tasks in the context of reasoning about actions in ther cellular domain. Others include prediction (hypothetical reasoning), explanation, and diagnosis. In the context of a cell, representing the cell behavior using actions and determining the effects and side effects of a drug (that manifests as a ligand) would correspond to prediction. Similarly, figuring out what interventions will change the cell behavior in a particular way corresponds to planning; determining the cause behind some unexpected observations about the cell corresponds to diagnosis. Since research on cell decoding is ongoing this project will have only partial information about the happenings inside a cell, and its reasoning schemes will need to be able to deal with incomplete information. In this way, this project ties into important goals in AI research on knowledge representation and reasoning. Although the existing research in knowledge representation and reasoning provides a good foundation for representing and reasoning about cell behavior, there are several new features in this domain that pose new research challenges. Tackling them is the main goal of this project. Challenges to be addressed include modeling triggering activity, notions of sensitization of receptors, and protein 'recruiting' proteins.