NECSI Resources

 International Conference on Complex Systems (ICCS2007)

Multi-level behaviours in agent-based simulation: colonic crypt cell populations

Chih-Chun Chen
Department of Computer Science, University College London

Sylvia Nagl
Department of Oncology and Biochemistry, University College London

Christopher Clack
Department of Computer Science, University College London

     Full text: PDF
     Last modified: October 8, 2007

Abstract
Agent-based modelling and simulation is now beginning to establish itself as a suitable technique for studying biological systems. However, a major issue in using agent-based simulations to study complex systems such as those in Systems Biology is the fact that simulations are `opaque'. While we have knowledge of individuals' behaviour through agent rules and have techniques for evaluating global behaviour by aggregating the states of individuals, methods for identifying the interactive mechanisms giving rise to this global behaviour are lacking. Formulating precise hypotheses about these multi-level behaviours is also difficult without an established formalism for describing them. The complex event formalism allows relationships between agent-rule-generated events to be defined so that behaviours at different levels of abstraction to be described. Complex event types define categories of these behaviours, which can then be detected in simulation, giving us computational method for distinguishing between alternative interactive mechanisms underlying a higher level behaviour. We apply the complex event formalism to an agent-based model of cell populations in the colonic crypt and demonstrate how competition and selection events can be identified in simulation at both the individual and clonal level, allowing us to computationally test hypotheses about the interactive mechanisms underlying a clone's success.

Appendices
      Clone population dynamics and example correlation analyses







Maintained by NECSI Webmaster Copyright © 2000-2007 New England Complex Systems Institute. All rights reserved.