Simulation Design
The simulator of cell expansion was made using the Pygame module of
python and is hosted on Zenodo and GitHub -
https://github.com/Sakib1418/Game-of-cells. The simulation was
designed to integrate with OpenAI gym , a collection of simulated
environments and associated toolkits to test and compare RL agent
algorithms. As the new gym environment was made, the Stable Baselines3
module was used on top of gym to explore current RL algorithms. The
properties of the actors (cells) attempt to simulate actual CAR T-cells,
for example movement and regeneration rate. Due to current lack of
measured parameters such as conversion probability on encountering a
bead, reasonable estimates are made in this initial work. All simulation
values and cell parameters are listed in Table 2. To observe agent
response with different cells, new cell types are conceptualized by
changing these cell-properties (Table 1). How these parameters are
formed into equations governing the fate of the cell and on the the
culture environment or simulation trajectory overall is detailed in the
game pseudocode (Supplement 2). Installation of the simulation-game,
data analysis and reproduction of the plots and usage are detailed in
the project GitHub repository.
Table 2: Parameters and their descriptions