Agent Based Immune System Simulator for "in machina" studies on human immunosenescene
Silvana Valensin*, Gianni Di Caro#, Claudio Franceschi* ^
*Department of Experimental Pathology, University of Bologna, Bologna, Italy
#I.R.I.D.I.A., Université Libre de Bruxelles, Bruxelles, Belgium
^I.N.R.C.A. - Italian National Research Center on Ageing, Ancona, Italy
ABISS has been developed envisaging the Immune System (IS) as a self organising system [1,2]. Immunosenescence [3-8] is a systemic process , that can be properly understood only taking into account the behaviour of the IS as a whole. Since we conceptualise aging, and consequently immunosenescence, as the partial and progressive disruption of structural equilibrium at various levels of sub-systems coordination, our aim is to address the study of the different levels of IS structural complexity as emerging from different levels of interactions of its cellular and molecular components. In agreement with our perspective, in developing ABISS we chose to adopt the modelling philosophy based on agents and the implementing method based on object oriented programming language (C++), and to introduce since the beginning an extended base of immunological knowledge .
The immunological knowledge base included in ABISS involves the following immune cells: T helper lymphocytes of types Th1 and Th2, cytotoxic T lymphocytes (CTL), B lymphocytes, NK cells, dendritic cells, macrophages, and the following molecules: soluble antibodies and cytokines. Generic somatic cells and antigens of various types, like self, viral and cellular antigens, are involved by immunological interactions. Interactions, which are based on condition/action rules, are regulated by the presence/absence of various additional molecules like receptors (TCR for T lymphocytes, antibodies for B lymphocytes), co-stimulatory molecules, ligands, surface markers, cytokine receptors, and MHC type I and II molecules.
Mechanisms like cell poiesis, proliferation, differentiation and death are taken into account, as well as the virus and bacteria life cycle. Interaction rules are many and complex. Here (figure 1) a sketch of the various conditions governing the changes of the functional status of the B cells is given. For sake of simplicity no specific immune organs or somatic districts are included.
All immune cells, soluble antibodies, antigens and somatic cells are modeled as agents and, more precisely, as reactive agents with an internal status. Since each of them has different specific properties the model is a Multi Agent System (MAS) based on heterogeneous agents. The IS, and consequently its multi agent model, is functionally, spatially, and temporarily distributed.
A vector of status is assigned to each agent, describing its internal status, to which many variables like functional status, age, presence of surface markers, and so on, contribute. Each agent undergoes over time an iterated process of local interactions with both, the other agents and the environment, resulting in status changes and actions, which are both performed according to the perceptions (sensory inputs) the agent receives from the environment (e.g. concentration of cytokines), to its actual internal status, and to that of the other agent it is interacting with. Thus, agent decisions (behaviour) depend upon their past actions, i.e. upon their own historical evolution in terms of internal status. Rules of interactions are defined in harmony with the immunological knowledge base.
The space in which agents are situated is divided in two different sub-spaces (figure 2), the immunological space and the somatic space, both tassellated and communicating between them in order to allow agent interactions. Interaction cycles happen at discrete time steps and are virtually parallel. At each time step, after the two cycles of interactions among the elements belonging to the same site (first cycle for both sub-spaces) and to corresponding sites of the two sub-spaces (second cycle), all agents but those representing somatic cells, can move to adjacent sites. Interactions within the same tassel, associated with the possibility to diffuse to adjacent sites, model the spatial and time proximity that influences the coordination and synchronisation of actions. A fundamental characteristic of ABISS is the flexibility, since it is designed with the purpose to give to the user a tool as much flexible as possible (flexibility meaning the possibility to freely choose which type of cells to include in the run and the value of many parameters), in order to specifically customize the simulation at different levels of complexity.
Up: The Approaches