Regulation

All currently implemented gene regulation models assume (discrete-space) pure jump process dynamics and (only) allow stochastically realizing concrete trajectories. For this purpose, they are first instantiated as Models.SciML.JumpModels containing Catalyst.ReactionSystems that are assembled according to various kinds of model definitions. This compilation process proceeds in multiple stages that recursively translate higher- to lower-level definitions, and each stage wraps its result in a Wrapped model along with its (intermediate) definition.

While most of this package assumes a dynamics interpretation in terms of jump processes and discrete counts, when the package is used as a library, the intermediate definitions are accessible (including the underlying Catalyst.ReactionSystem and JumpProcesses.JumpSystem). It is therefore possible, for example, to use these objects for static analysis with other SciML tooling or to simulate trajectories from an ODE relaxation. Given any Wrapped model object, its definition property holds just that, and accessing the model property peels off one layer. For example, given a JSON model specification

d = """
    {"{regulation/kronecker}": {
        "seed": "seed",
        "base_rates": {"\$": ["defaults", "gene", "base_rates"]},
        "activation": {
            "adjacency": {
                "initiator": [[0.7, 0.6], [0.4, 0.2]],
                "power": 3
            },
            "at": ["LogNormal", 2.0, 1.0]
        },
        "repression": {
            "adjacency": {
                "initiator": [
                    [0.7, 0.6],
                    [0.5, 0.2]
                ],
                "power": 3
            },
            "at": ["LogNormal", 2.0, 1.0]
        }
    }}
"""

nothing

we have:

julia> model = Models.parse(d);

julia> model isa Models.Wrapped
true

julia> typeof(model.definition)
GeneRegulatorySystems.Models.KroneckerNetworks.Definition

julia> typeof(model.model.definition)
GeneRegulatorySystems.Models.V1.Definition

julia> typeof(model.model.model.definition)
Catalyst.ReactionSystem{Catalyst.NetworkProperties{Int64, SymbolicUtils.BasicSymbolic{Real}}}

julia> typeof(model.model.model.model)
GeneRegulatorySystems.Models.SciML.JumpModel

julia> typeof(model.model.model.model.system)
ModelingToolkit.JumpSystem{RecursiveArrayTools.ArrayPartition{Any, Tuple{Vector{JumpProcesses.MassActionJump}, Vector{JumpProcesses.ConstantRateJump}, Vector{JumpProcesses.VariableRateJump}, Vector{Symbolics.Equation}}}}

Higher-order model blueprints such Differentiation are expressed in terms of a common first-iteration model of simplified gene regulation defined in the Models.V1 module. Some of these models (named "templates", currently Kronecker-linked networks and Random differentiation) are stochastic and only become concerete model definitions (i.e., their structure deterministic) by affixing a seed.

V1 model

GeneRegulatorySystems.Models.V1 — Module

Contains components to build SciML.JumpModels for gene regulation.

These models can be constructed by build, either directly (from a Definition or from the corresponding JSON specification), or indirectly from a model template such as KroneckerNetworks.Definition or Differentiation.Definition via their respective build functions.

The regulation models require setting up the initial state to include required molecular species (polymerases, ribosomes and proteasomes) before they can be invoked. If a model is specified as part of a Schedule and therefore the defaults are available, this can be achieved by prefixing the model specification with an Instant adjustment step such as:

{"{add}": {"$": ["defaults", "bootstrap"]}}

See examples/specification/simple.schedule.json for an example.