Agent-based model of CAR T-cell therapies in dish and tissue contexts

• Code structure overview
• Building from source
• Running the model
• Setup file structure
  • <simulation> tags
    • <profilers>
    • <checkpoints>
  • <agents> tags
    • <populations>
    • <helpers>
  • <environment> tags
    • <globals>
    • <components>

Code structure overview

carcade.jar is a compiled jar of the model, with the required library MASON and other libraries (listed below) included.

The src/ directory contains all source files for the model.

The docs/ directory contains class documentation (javadoc), sample input XML files (sample_input), and sample output JSON files (sample_output). The documentation can be viewed in a web browser at https://bagherilab.github.io/CARCADE/javadoc/index.html. There is also a README file in the docs/ directory with more details on the sample input and output files.

Building from source

The model can be built from source. The following libraries are required:

• mason.19.jar
• flatlaf-0.33.jar
• vecmath.jar
• j3dcore.jar
• j3dutils.jar

Running the model

The model can be run either command line or GUI (which can use a 2D rendering).

When using command line, first navigate to the directory holding this file. Then:

java arcade.Main XML [(--vis)] [(-v|--view) VIEW]

    XML
        Path to .xml setup file
    [(--vis)]
        Flag for running with GUI
    [(-v|--view) VIEW]
        Selects the GUI view: 2D (default).

For example, to run the model with setup file setup.xml in GUI mode with a 2D view:

java arcade.Main setup.xml --vis --view 2D

Note that running the model with visualization (either directly from the arcade.jar or using the --vis flag) is significantly slower than without visualization.

You can run also the model from a .jar using the carcade.jar file. For example, running the simulation with setup file setup.xml, without a GUI, and using the .jar file:

$ java -jar carcade.jar setup.xml

Setup file structure

The model uses an .xml file that describes the setup of the simulations, agents, and environments. Attributes in brackets ([]) are optional. Values in ALL CAPS should be replaced with actual values. The basic structure of the file is:

<set prefix="PREFIX" path="PATH">
    <series name="NAME" start="START" end="END" days="DAYS">
        <simulation type="TYPE" radius="RADIUS" height="HEIGHT" margin="MARGIN">
            <!-- SIMULATION TAGS HERE -->
        </simulation>
        <agents initialization="INIT" plate="PLATE">
            <!-- AGENTS TAGS HERE -->
        </agents>
        <environment coordinate="COORD">
            <!-- ENVIRONMENT TAGS HERE-->
        </environment>
    </series>
    ...
</set>

A group of simulations series is called a set. The setup XML can only include one set.

• path = absolute path to the output directory
• [prefix] = prefix for output files

A simulation series is a group of simulations that only differ in the random seed. Each <series> includes <simulation>, <agents>, and <environment> tags that define various parts of the series. A set can include multiple <series> tags.

• name = name of the simulation series for output files
• [start] = (integer) starting seed (default = 0)
• [end] = (integer) ending seed (default = 1)
• days = (integer) number of days for each simulation

The simulation tag describes the size and type of the simulation. Nested tags include various profilers and checkpointing for the simulation.

• type = simulation type (only growth is currently supported)
• [radius] = (integer) radius of simulation (default = 34)
• [height] = (integer) height of simulation (default = 1)
• [margin] = (integer) margin of simulation (default = 6)

The agents tag describes the agents initialization and type of plating. Nested tags include definitions of the various cell populations to include (as well as cell modules and/or population parameters) and any helpers.

• initialization = (integer) radius to which cell agents are seeded
  • use 0 for no cells
  • use FULL to seed up to the RADIUS of the simulation
• plate = plating configuration for how cell agents are seeded
  • use dish for a dish configuration in which cell agents are plated randomly across the entire simulation
  • use spheroid for configuration in which tissue cells are seeded in a tumor spheroid starting from the center and moving outward until all cells are placed (default)

The environment tag describes the coordinate system of the environment. Nested tags include environment parameters and components.

• coordinate = coordinate system for the simulation (hex = hexagonal)

<simulation> tags

<profilers>

Profilers save the simulation state to .json files at the selected interval. Different profilers will save different types of information. These are not used when the model is run as a GUI.

<profilers>
    <profiler type="growth" interval="INTERVAL" suffix="SUFFIX" />
    <profiler type="parameter" interval="INTERVAL" suffix="SUFFIX" />
    <profiler type="graph" interval="INTERVAL" suffix="SUFFIX" />
    <profiler type="lysis" interval="INTERVAL" suffix="SUFFIX" />
    ...
</profilers>

• growth saves a profile of cell state (including volume and cycle length) and a span of concentrations at the given interval
  • interval = (integer) minutes between each profile
  • [suffix] = appended to the output file name before the extension
• parameter saves a profile of cell parameters at the given interval
  • interval = (integer) minutes between each profile
  • [suffix] = appended to the output file name before the extension
• graph saves a profile of graph vasculature at the given interval
  • interval = (integer) minutes between each profile
  • [suffix] = appended to the output file name before the extension
• lysis saves a profile of cell lysis (including type and time of death) at the given interval
  • interval = (integer) minutes between each profile
  • [suffix] = appended to the output file name before the extension

<checkpoints>

Checkpoints save certain parts of the simulation to a .checkpoint file, which can later be loaded. These are included when the model is run as a GUI, although the seed may not match. They can either be class SAVE or LOAD, depending on if the checkpoint is being saved from or loaded to the simulation.

<checkpoints>
    <checkpoint type="cells" class="CLASS" name="NAME" path="PATH" day="DAY" />
    <checkpoint type="graph" class="CLASS" name="NAME" path="PATH" day="DAY" />
    ...
</checkpoints>

• cells saves a checkpoint of all the cell agents OR loads a checkpoint of cell agents
  • name = name of the checkpoint file
  • path = path to the checkpoint file directory
  • [day] = (integer) day that the checkpoint is to be saved or loaded on
• graph saves a checkpoint of the graph structure (only use with the graph component)
  • name = name of the checkpoint file
  • path = path to the checkpoint file directory
  • [day] = (integer) day that the checkpoint is to be saved (graph checkpoints can only be loaded when the simulation is initialized, i.e. day = 0)

<agents> tags

<populations>

Populations define the cell populations included in the simulation. Within each population, you can also optionally define variables and modules.

<populations>
    <population type="TYPE" fraction="FRACTION">
        <variables>
            <variable id="PARAMETER_NAME" value="VALUE" scale="SCALE" />
            ...
        </variables>
        <modules>
            <module type="TYPE" complexity="COMPLEXITY" />
            ...
        </modules>
    </population>
    ...
</populations>

The population tag describes the type and fraction of a cell population included in the model.

• type = cell type (H = healthy tissue cell, C = cancerous tissue cell, S = cancer stem cell, 4 = CD4+ CAR T-cell, and 8 = CD8+ CAR T-cell)
• fraction = (number) fraction of this population in the initialized cells, total fractions across all populations should sum to 1.0

The variable tag lists parameters for the population that are either set to a new value or scaled from the default value. Unless modified, default values are used for all cell parameters. Either or both value and scale attributes can be applied, with value applied first.

• id = name of the parameter to be modified
• [value] = (number) new value for the parameter
• [scale] = (number) scale value of the parameter

The module tag lists the tissue cell modules. When undefined, complex metabolism and complex signaling are used by default. CAR T-cell agents do not need specified modules, as they use default metabolism modules for CAR T-cell agents and subtype-specific inflammation modules for CAR T-cell agents.

• type = module type, currently support for metabolism and signaling
• complexity = complexity of the module (C = complex, M = medium, S = simple, R = random)

<helpers>

Helpers define the various helper agents included in the simulation. Depending on the type, different attributes are required.

<helpers>
    <helper type="convert" delay="DELAY" population="POPULATION" />
    <helper type="insert" delay="DELAY" populations="POPULATIONS" bounds="BOUNDS"/>
    <helper type="wound" delay="DELAY" bound="BOUNDS" />
    <helper type="treat" delay="DELAY" dose="DOSE" ratio="RATIO" />
    ...
<helpers>

• convert changes the cell in the center of the simulation to a cell of the given population after delay time has passed
  • population = (integer) zero-indexed population number for the cell to convert to
  • delay = (integer) minutes before convert occurs
• insert adds a mix of cells of the selected populations at a radius of bounds*RADIUS after delay time has passed
  • populations = (comma-separated integers) list of zero-indexed population numbers to add
  • bounds = (number) fraction of total simulation radius to add cells
  • delay = (integer) minutes before insert occurs
• wound removes all cells within a radius of bounds*RADIUS after delay time has passed
  • bounds = (number) fraction of total simulation radius to remove cells
  • delay = (integer) minutes before wound occurs
• treat adds the dose of CAR T-cells at CD4+:CD8+ ratio ratio throughout the simulation with bias towards cancerous cells after delay has passed, CAR T-cell populations for treatment must be initialized at 0.0 fractions in <populations> tags
  • dose = (integer) total quantity of CAR T-cells to add to the simulations
  • ratio = (decimal:decimal) fraction of CD4+ and CD8+ CAR T-cells, respectively, where fraction for each is multiplied by the dose to determine the total quantity per type, and total fractions across both populations should sum to 1.0
  • delay = (integer) minutes before treatment occurs

<environment> tags

<globals>

Globals define environment parameters that are changed, analogous to the population-specific variables. Note that some global parameters may only apply to certain components. Unless modified, default values are used for global parameters. Either or both value and scale attributes can be applied, with value applied first.

<globals>
    <global id="PARAMETER_NAME" value="VALUE" scale="SCALE" />
    ...
<globals>

• id = name of the parameter to be modified
• [value] = (number) new value for the parameter
• [scale] = (number) scale value of the parameter

<components>

Components define various components included in the simulation. Depending on the type, different attributes are required. At least one sites type component must be specified.

<components>
    <component type="sites" class="source" x="X" y="Y" z="Z" />
    <component type="sites" class="pattern" />
    <component type="sites" class="graph" layout="LAYOUT"
        left="LEFT" right="RIGHT" top="TOP" bottom="BOTTOM" />
    ...
<components>

• sites/source uses constant source sites, defined by x, y, and z
  • [x], [y], [z] = single number, start:end range, or start:increment:end interval of x, y, or z coordinates for the sources
  • default * indicating the entire range
• sites/pattern uses pattern-based sites
• sites/graph uses graph-based sites
  • [layout] = structure of the vasculature roots (if equal to *, then layout matches that of the pattern-based sites but in graph form; if empty, then a graph much be loaded from a checkpoint)
    • S = single roots, defined as comma-separated list of (###)(AV) where (###) is the percent distance along the border and (AV) indicates artery (A) or vein (V)
    • A = alternating roots, defined as number of sites along the given border (which will alternate between artery and vein)
    • R = random roots, defined as number of sites along the given border (which will be randomly positioned and randomly assigned as artery or vein)
    • L = line roots, defined as comma-separated list of (###)(AV)(000) where (###) is the percent distance along the border, (AV) indicates artery (A) or vein (V), and 000 is the percent distance covered by the line
  • [left], [right], [top], and [bottom] = root locations defined depending on the the layout type
  • by default, the graph sites use realistic hemodynamics, use type="graph.simple" to instead use a simplified version
• degrade degrades the cell wall of vessels located where there are non-healthy cell agents (only use with site/graph)
  • interval = (integer) minutes between degradation
• remodel remodels the cell wall and vessel radius based on stress and flow (only use with site/graph)
  • interval = (integer) minutes between remodeling