A multi-agent system for the quantitative simulation of biological networks

  • Authors:

  • Salim Khan;Ravi Makkena;Foster McGeary;Keith Decker;William Gillis;Carl Schmidt

  • Affiliations:

  • University of Delaware, Newark, DE;University of Delaware, Newark, DE;-;University of Delaware, Newark, DE;University of Delaware, Newark, DE;University of Delaware, Newark, DE

  • Venue:
  • AAMAS '03 Proceedings of the second international joint conference on Autonomous agents and multiagent systems
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

We apply the multi-agent system (MAS) platform to the task of biological network simulation. In this paper, we describe the simulation of signal transduction (ST) networks using the DECAF [9] MAS architecture. Unlike previous approaches that relied on systems of differential equations (DE), the distributed framework of MAS scales well and allows us to model large, highly interconnected ST pathways. This scalability is achieved by adopting a hybrid strategy that factors macro-level measures, such as reaction rateconstants, to calculate the stochastic kinetics at the level of individual molecules. Thus, by capturing the ST domain at an intermediate level of abstraction, we are able to retain much of the granularity afforded by a purely individual-based approach. The task distribution within a MAS enables us to model certain physical properies, such as diffusion and subcellular compartmentalization, which have proven to be difficult for DE systems. We demonstrate that large-grained agents are well suited to maintaining interal state representations and efficient in computing reactant concentration, both of which are vital considerations in modeling the ST domain. In our system, a molecular species is modeled as an individual agent with hierarchical task network structures to represent self- and externally-initiated reactions. An agent's identity is determined by a rule file (one for every participating molecular species) that specifies the reactions it may participate in, as well as its initial concentration. Reactions within the system are actuated by inter-agent communication. We present results from modeling the well-studied epidermal growth factor (EGF) pathway, demonstrating the viability of MAS technologies as a simulation platform for biological networks.