Courtesy of B. Gallagher and Thierry Emonet
Agent-based simulation of 1,080 cells in a 3D medium with a vertical aspartate gradient, shown at 54, 150, and 400 seconds after simulation start. 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.
A new, single-cell computational model developed by scientists at the University of Chicago and Argonne National Laboratory borrows a technique used in the social sciences to digitally study how random molecular events within a cell influence its behavior. In a proof-of-principle study, Thierry Emonet and colleagues showed that their program, AgentCell, accurately simulated chemotaxis in 1,000 individual
The authors plan to use the model to study interactions between cells. "With AgentCell, you can go from the molecular events inside a single cell, to the behavior of that single cell inside of a 3D environment, and then from the communication between those cells, you can also involve the dynamic of the population," says Emonet, a research scientist in Philippe Cluzel's laboratory at the University of Chicago.
While other single-cell modeling programs (for example, E-Cell and the Virtual Cell) aim to simulate all of the systems within a single cell, AgentCell uses high-level rules to model one well-understood biological system – in this case, chemotaxis. "This will be a very valuable complement to 'bottom-up' modeling tools like StochSim or Virtual Cell, which focus on detailed biophysical mechanisms underlying cell biological events," cell biologist and Virtual Cell developer Les Loew of the University of Connecticut Health Center, in Farmington, writes in an E-mail.
In agent-based modeling and simulation (ABMS) – an approach frequently used to simulate social and economic systems – autonomous "agents" (in this case, individual bacterial cells) make decisions according to a set of defined rules that govern their behavior. In AgentCell, these rules derive from the group's prior experimental observations of the chemotaxis system in
North says users can easily add new rules or modify existing ones, or do sensitivity analysis to determine which rules have the most effect on cell behavior. "We can actually turn individual rules on and off in the system in any combination, and do tests to see which of the rules actually are making the biggest difference to the outcome," he says.
Furthermore, the system is modular. "The software is built in such a way that we can actually update future versions without modifying drastically the code," says Cluzel. And because AgentCell is open-source, users in the chemotaxis research community can help improve the code. According to Emonet, the software was made available in mid-August.
AgentCell's modularity stems in part from its use of StochSim, an E-Cell precursor that has been used for several years to model chemotaxis in single cells. "AgentCell includes that code inside each cell. So basically when you run a simulation, you have 1,000 cells in a population, and inside each one of those cells you have an independent simulation of that StochSim package, that simulates the biochemistry," says Emonet.
One possible future application for AgentCell is modeling the interactions between a cell and its environment. "One can imagine ... that you could model the interaction of the cell to the host in the actual environment of the host," says computational biologist Christopher Rao of the University of Illinois.
Emonet, though, says the next step is to use the model to investigate quorum sensing – for example, determining if cells communicate with each other or act autonomously when searching for food. "Certain bacteria act individualistically for a while, and then suddenly they switch mode and they become very social and they behave as a social group. And then they go back to individual behavior. And all that can be studied with this kind of code," he says.
But StochSim codeveloper Tom Shimizu, a postdoctoral fellow in Howard Berg's laboratory at Harvard University, says investigating intracellular events may be computationally demanding. "What they've done so far with AgentCell is show very nicely how you could scale up the kind of computations that were being done for intercellular pathways in single cells, to doing a large number of them in parallel using this framework," says Shimizu. Incorporating intracellular communication requires the entire code to be parallelized so that all processes can talk to one another, Shimizu explains. "Even if you did that implementation, the cost of communicating makes the whole program run a lot slower. I think that would be a big challenge," he says.
Coauthor Charles Macal of Argonne National Laboratory agrees that efficient parallelization is the next hurdle. "This will be an area of intense investigation and experimentation for us starting immediately," he says.