Because of the highly regulated nature of AAM, the lung can be considered as an immunosuppressive organ for respiratory patho- gens. However, as infection progresses in the lung, another population of macrophages, known as classically activated macro- phages (CAM), enters; these cells are typically activated by IFN-γ. CAM are far more effective than AAM in clearing the microbial load, producing proinflammatory cytokines and antimicrobial defense mechanisms necessary to mount an adequate immune response. The present work is concerned with determining the first time when the population of CAM becomes more dominant than the population of AAM. This proposed “switching time” is explored in the context of Mycobacterium tuberculosis infection. Methods: A mathematical model based on a system of ordinary differential equations is developed, which describes the interactions among cells, bacteria, and cytokines involved in the activation of both AAM and CAM during a tuberculosis infection in the lung. Sensitivity analysis is preformed on specific model parameters to determine those parameters that play a major role in affecting the switching time and related bacterial loads. Because IFN-γ is a necessary cytokine for the classic activation of macrophages, we also use our model to determine how early introduction of IFN-γ may act as a therapeutic agent to alter the switching time and lessen the bacterial loads. Results: During the course of a simulated Mycobacterium tuberculosis infection, the model predicts a switching time of 50 days and shows that the AAM not only ineffectively deal with the bacteria but also prevent early recruitment of necessary effector cells, positioning their bacterial opponent at an unfair advantage. This immune battlefield may also negatively influence vaccine strategies in the lung microenvironment. Hence, if the switching time could be altered to occur earlier in the response, then, theoretically, tuberculosis therapies along with a more robust immune system may clear the disease more effectively because a reduced switching time may imply reduced peak bacterial loads. Treatment simulations involving IFN-γ therapy reduce the switch- ing time to 34 days and reduce peak and residual bacterial loads; however, the bacteria are not eradicated at 100 days. Conclusions: Generally speaking, our results suggest that a reduction in the switching time correlates with lower peak and residual bacterial loads and also that therapeutic strategies should aim to reduce bacterial numbers but not reduce the signaling to downstream mediators. In addition, our results imply that IFN-γ alone is not sufficient to mediate a sterilizing immune response for tuberculosis. In agreement with this finding, clinical trials with IFN-γ therapy, to date, have not proven to be highly efficacious, especially in the long term. doi:10.1016/j.jcrc.2009.06.020 A parallel implementation of an agent-based modeling platform with application in modeling calcium releases in cardiomyocytes Maxim Mikheev a,e , Alexey Solovyev b , Anna Maltsev c , John Bartels f , Steven Chang f , Qi Mi d,e , Yoram Vodovotz a,e a Department of Surgery, University of Pittsburgh b Department of Mathematics, University of Pittsburgh c Department of Mathematics, California Institute of Technology d Department of Sports Medicine and Nutrition, University of Pittsburgh e Center for Inflammation and Regeneration Modeling (CIRM) at the McGowan Institute of Regenerative Medicine, University of Pittsburgh f Immunetrics, Pittsburgh, PA Objectives: We recently developed a novel agent-based modeling platform: SPARK (Simple Platform for Agent-based Representation of Knowledge) for multiscale biomedical agent-based modeling (ABM). The initial version of SPARK used a sequential execution algorithm that restricts the size of simulation and that cannot take full advantage of the computational power of multicore PCs and cluster machines. Herein, we report on a parallel implementation of SPARK. We demonstrate an initial application in modeling Ca 2+ release in cardiomyocytes. Methods: We applied the Java Parallel Arrays library (jsr166y) to implement parallelization on a single multicore computer. This library implements the Open Multi-Processing (OpenMP) API, which supports multi-platform, shared memory, and multiprocessing programming. Both agent actions and data layer computations were parallelized. To implement parallelized SPARK ABM's on a cluster, we divided the space into smaller parts and assigned each part to a particular machine in a cluster. Because agents occupy a particular part of the space, we distributed the agents among machines in a spatial manner. For cluster synchronization, we used the P2P-MPI imple- mentation of the Message Passing Interface (MPI) protocol. Results: We tested the parallelized SPARK successfully on an Intel- based IBM computing cluster as well as multicore, Intel-based desktop computers. We used the parallelized SPARK framework to extend a mathematical model of stochastic local Ca 2+ release in cardiomyocytes (Bioph J 2007, Abstract, p344a) as an ABM. This ABM, which can be used for both basic studies and prediction of response of cells to drug treatment, was tested by execution of 5000 steps on a dual-core machine. We found that execution time was 184 vs 357 seconds for 2 cores vs 1 core, respectively. Conclusions: We developed a parallelized version of SPARK, which allows the much faster execution of ABMs. Future studies will involve ABMs of inflammation and healing. doi:10.1016/j.jcrc.2009.06.021 Simulation of lung alveolar type II epithelial wound healing in vitro Sean H.J. Kim a , C. Anthony Hunt a,b a UCSF/UC Berkeley Joint Graduate Group in Bioengineering, University of California, Berkeley b Department of Bioengineering and Therapeutic Sciences, University of California, San Francisco Objectives: In vitro wound healing studies have shown that alveolar type II (AT II) cells in 2-dimensional environment recapitulate features of early-stage alveolar epithelial repair after acute lung injury. Similar to in vivo repair, cell migration and spreading are primarily responsible for initial wound closure; cell proliferation plays no significant role. Mechanisms that enable and regulate the process in vitro and in vivo are unknown. Our objective is to explore, characterize, and understand plausible mechanisms, using an agent- based, discrete event simulation approach. Methods: The approach evolved from earlier studies of modeling cell systems, leveraging new, object-oriented, executable biology capabilities. It is based on a theory that when 2 model systems—in vitro cell cultures and an in silico analogue—are composed of components for which similarities can be established, and the 2 systems exhibit multiple attributes that are similar, then there may also be similarities in the generative mechanisms responsible for those attributes. Using the approach, we created multiagent, in silico e21 Meeting Abstracts for the Society for Complexity in Acute Illness (SCAI)