AgentCell (Digital E. Coli)

T. Emonet, C. Wickersham & P. Cluzel (Institute for Biophysical Dynamics, University of Chicago)
C. Macal, M. North (Center for Complex Adaptive Agent Systems Simulation, Argonne National Laboratory)
B. Gallagher (ASC Flash Center Visualization Group, University of Chicago)

This work is partially supported by joint research funding under H.28 of the U. S. Department of Energy Contract W-31-109-ENG- 38. It is an ongoing collaboration between P. Cluzel's lab (UofC) and CCAASS (Argonne National Laboratory)

Introduction

In recent years, single-cell biology has focused on the relationship between the stochastic nature of molecular interactions and variability of cellular behavior. To describe this relationship, it is necessary to develop new computational approaches at the single cell level.

We have developed AgentCell, a model using agent-based technology to study the relationship between stochastic intracellular processes and behavior of individual cells. As a test-bed for our approach we use bacterial chemotaxis, one of the best-characterized biological systems. In this model, each bacterium is an agent equipped with its own chemotaxis network, motors and flagella. Swimming cells are free to move in a 3D environment. Digital chemotaxis assays reproduce experimental data obtained from both single cells and bacterial populations.

(a) Chemotactic swimming cell exposed to a linear gradient of chemo-attractant (shaded background). When the motors rotate CCW (see main text) the flagella form a corkscrew bundle and the cell runs. When the motors rotate CW, the flagella bundle flies apart and the cell tumbles erratically. (b) Schematic diagram of the chemotactic pathway. Receptor complexes consist of three Tar receptor dimers (T), two CheW (W), and two CheA (A). The activity of aspartate (a) receptor complexes is modulated by methylation (m) and demethylation respectively catalyzed by CheR (R) and CheB (B). The autophosphorylation of CheA (A), a histidine kinase, is controlled by the activity of the receptors. CheA catalyzes the phosphorylation of the response regulator CheY (Y). When CheY is phosphorylated (Yp), it binds to the cytoplasmic domain of the bacterial motor and induces tumbles. Dephosphorylation of CheY-P (Yp) is catalyzed by the phosphatase CheZ (Z).

Results and animations

First results are summarized in the article: "AgentCell: a digital single-cell assay for bacterial chemotaxis", Bioinformatics (2005), 21(11), 2714-2721 (online)
View a real E. coli bacterium swimming in a 2D environment ( avi movie by P. Cluzel)
Simulation of 1166 cells in a 3D medium without attractant. The cell diffuse away from the initial location: the squared average distance from the initial position grows linearly with time (3D rendering by B. Gallagher) (avi)
View (avi) a digital E. coli bacterium swimming in a vertical gradient (10^-8 M/micrometer) of aspartate. E. coli uses the chemotaxis sensory system to control its flagellar motors and move towards the source of nutrient (here the concentration of nutrient increases with the height z). The movie shows the intracellular level of response regulator protein CheY-P (CheY-P induces tumbling) and the receptor occupancy by ligand molecules (aspartate). Tumbling events are in red.
1080 cells in a 3D medium with a vertical gradient of aspartate (10^-8 M/micrometer). 540 cells are sensitive to aspartate (green) and 540 cells are not sensitive (red). To illustrate the complicated trajectory of cells, the trace of two typical cells is shown. The right panel shows the chemotactic response of the population: the total number of cells that have reached the height z=1mm. After 400 secs, about one half of the red cells have reached regions of high attractant concentrations causing their receptors to saturate. From that time point on, the red population diffuses away at the same rate as the green cells (figure).
Movie files: ( 800x600 avi 112 Mb) ( 1024x768 avi 169 Mb)

To download high resolution (printable) versions of these images click here.

Software and bug fixes

AgentCell is now fully compatible with the latest version of StochSim (StochSim-1.6_core_2007JUL05.tar.gz NOT to be mixed up with the 1.6beta version of StochSim). Our changes to StochSim have now been merged into the main development trunck of StochSim. The latest version of StochSim on the sourceforge server is now fully compatible with this new version of AgentCell. StochSim is not distributed with AgentCell anymore. It is available on sourceforge. We still provide the StochSim file here for convenience. AgentCell uses parts of Repast as a layer to provide agent-based capabilities. Repast is distributed under a BSD license and is included in the AgentCell distribution. AgentCell is distributed under the GPL license. AgentCell includes Xerces XML software developed by the Apache Software Foundation (http://www.apache.org/). Please see the "licenses" directory for the licensing details of AgentCell and the software distributed with AgentCell.

AgentCell Version 2.0 (download) (README)

This new version of AgentCell can now handle many cells in the same VM. It is also fully compatible with the latest version of StochSim that can be found on sourceforge.

AgentCell Version 1.2 (download)

This version still uses version 1.4 of StochSim but with some important bug fixes:

Fixed the time counter overflow: Due to use of double in stochsim to store the time and conversion between double and long, the time variable was overflowing at some point causing the cells to become "blind" or insensitive after about 400 seconds of simulation. The sudden change in the slope of the chemotactic response of wild type cells in figure 4 of Bioinformatics, 21, 2714-2721 (2005) at time t=400 secs is not due to the saturation of the receptors but is an artifact. With this bug fixed the wild type cells continue their strong chemotactic response until the end of the simulation at 1000 sec, which is good news. If you were using AgentCell for simulations shorter than 400 seconds or without ligand in the external medium your result should not be affected. This bug has been fixed using a quick fix. A more permanent solution is being implemented in stochsim: use of exact integer arithmetic to keep track of the number of time steps.
When defining more that 2 states for the receptors the updating of receptor states as the cell was moving was being ignored. That was a bug in the wrapping of stochsim. Now one can define receptors with as many states as wanted. Notice that the simulations in the examples and in the Bioinformatics 21, 2714-2721 (2005) paper only used 2 receptors with two states (number of dynamic values defined in Stochsim) and therefore these results were not affected by this bug.

AgentCell Version 1.0 (download)

Original version of AgentCell used in Bioinformatics 21, 2714-2721 (2005). Implementing the Network class with StochSim required the accessing of the internal time loop of StochSim in order to feed or extract information at the time intervals required by the agent-based simulation. For this purpose we modified StochSim to appropriately pause and then continue as required by Repast. This new version of StochSim eventually will become the main branch of the StochSim development tree (in version 1.4 of AgentCell)

Because of the use of global variables in StochSim 1.4, only one instance of StochSim 1.4.1 can be used in each Java VM. In practice we run many independent AgentCells on a cluster, one AgentCell per node (see explanations in the paper). Our goal however is to have cells interacting with each other. To that end Tom Shimizu and Michael North have modified StochSim 1.4.1, removing all global variables and therefore allowing many instances of the StochSim class within one AgentCell simulation. This new version of StochSim is currently being tested and will be released with the next version of AgentCell.

This version contains two important bugs that are fixed in version 1.2

Download

The code is available for download from SourceForge. To install download the distribution (zip file) from SourceForge, unpack and follow the instructions in the README file.

Getting involved

AgentCell is an open source project. Any contribution is welcome. To get involved, subscribe to the AgentCell-developer mailing list below.

Mailing lists

User questions should go to agentcell-interest@lists.sourceforge.net. To subscribe click here.
Developer questions should go to agentcell-developer@lists.sourceforge.net. To subscribe click here.

License

AgentCell is free software distributed under the GPL License. It includes other software such as (Repast, StochSim, ...). Licenses for each package can be found in the licenses directory within the distribution package. Repast is licensed under a BSD license. StochSim 1.4.1 (the modified 1.4 version) is included with AgentCell and is distributed under the LGPL license. AgentCell includes Xerces XML software developed by the Apache Software Foundation (http://www.apache.org/).