Link to paper
The full paper is available here.
You can also find the paper on PapersWithCode here.
Abstract
- Regulatory networks depict interactions between molecules in a biochemical system
- Signed graphs and signed functors are used to model and describe regulatory networks
- Functorial mappings are established between regulatory networks and other mathematical models in biochemistry
- Reaction networks modeled as Petri nets with signed links can be used to define a physical mechanism underlying a regulatory network
- Regulatory networks can be associated with a Lotka-Volterra system of differential equations
Paper Content
Introduction
- Living cells are made up of genes, proteins, and RNA molecules
- These molecules interact in complex ways to sustain the cell
- Regulatory networks are a directed graph that represent these interactions
- Edges are labeled with a positive or negative sign to indicate if the interaction is activating or inhibiting
- Network motifs are simple patterns that recur frequently in regulatory networks
- Category theory is a natural tool to study regulatory networks
- Sign-preserving functors can express indirect occurrences
- Regulatory networks can be viewed as signed graphs and signed categories
- Functorality is a safeguard for model transformation
- Regulatory networks can be connected to biochemical reaction networks
- Functorial assignment of continuous dynamics to regulatory networks is possible
- Lotka-Volterra systems of equations can be used to model complex biological systems
- Category theory is used to extend the constructions from closed systems to open ones
- Double categories are used to compose open systems together
Qualitative analysis: motifs and mechanisms
Regulatory networks as signed graphs
- Definition of a graph: a functor from Sch(Graph) to Set with two parallel morphisms
- Graph consists of a set of vertices and edges, and functions assigning source and target to each edge
- Signed graph is a graph with a function assigning a sign to each edge
- Regulatory networks are signed graphs with vertices representing components and edges representing interactions
- Sign-valued matrices are a special case of signed graphs
- Morphisms of signed graphs can embed or collapse multiple vertices/edges
- Signed graphs form a well-behaved category
- Signed graph morphisms factor essentially uniquely as an epimorphism followed by a monomorphism
Refining regulatory networks using signed categories and functors
- Signed morphisms do not capture the idea of refining regulatory networks
- Signed categories are needed to express refinement
- Signed categories have objects and morphisms with signs
- Morphisms of signed categories are called signed functors
- Signed functors preserve signs
- Signed path categories are generated by signed graphs
- Signed functors between signed graphs are determined by morphisms of signed graphs
- Positive autoregulation is an example of a network motif
- Instances of motifs are monic signed functors
- A functor maps regulatory networks to instances of motifs
- A commutative square of functors is used to extend the signed path category functor
Mechanistic models as petri nets with links
- Regulatory networks summarize how components of a complex biochemical system interact
- Regulatory networks include only a subset of the system’s components
- Regulatory networks do not model individual reactions and processes, only pairwise interactions
- Regulatory networks are not fully mechanistic models
- Mechanistic models in biochemistry model individual reactions
- Petri nets with signed links have two types of vertices and signed edges
- Signed edges represent reactions with multiple inputs/outputs, consumption, and promotion/inhibition
- Petri nets with signed links can be approximated as signed graphs
- A mechanistic model for a regulatory network is a Petri net with signed links and a monic signed functor
Parameterized dynamical systems
- Baez and Pollard extended the mass-action model of reaction networks to a functor from the category of Petri nets with rates into a category of dynamical systems
- Rate coefficients are often unknown and must be extracted from existing literature or estimated from experimental data
- The dynamics functor is nearly identical to Baez-Pollard’s
- The dynamics functor is the main building block in constructing a category of parameterized dynamical systems
- Many dynamical models depend linearly on their parameters
- The law of mass action and Lotka-Volterra equations are linear in the rate and affinity parameters
- To express important physical constraints and to define a semantics for signed graphs, the dynamical system and its parameters are restricted to be nonnegative
- There is a functor Dynam + : FinSet → Con that sends a finite set S to the conical space of essentially nonnegative, algebraic vector fields
- A conically parameterized nonnegative dynamical system consists of finite sets P and S together with a conic-linear map v : R P + → Dynam + (S)
- The categories of linearly and conically parameterized dynamical systems are finitely cocomplete
- The initial linearly parameterized dynamical system has no parameter variables, no state variables, and the unique (trivial) vector field on the zero vector space
- The coproduct of two linearly parameterized dynamical systems has parameter variables P 1 + P 2 , state variables S 1 + S 2 , and parameterized vector field
The lotka-volterra dynamical model
- A Lotka-Volterra system with n species has a vector field with state vector x ∈ R n and parameters ρ ∈ R n and β ∈ R n×n
- The parameter ρ i sets the rate of growth or decay for species i
- The parameter β i,j defines a promoting or inhibiting effect of species j on species i
- A functor from finite graphs to linearly parameterized dynamical systems gives a semantics for unlabeled graphs
- A functor from finite signed graphs to conically parameterized nonnegative dynamical systems gives a semantics for regulatory networks
- The functor LV preserves finite colimits
- The morphism LV(p) sends the parameterized dynamical system with state variables {R, * } and parameters ρ, β ∈ R {R, * } +
- The first way sets the system’s coefficients equal to sums of the former’s coefficients
- The second way substitutes x * for each x i , i ∈ S, in the first system and then takes the vector field v * to be the sum of the v i ’s
Composing lotka-volterra models
- Extending Lotka-Volterra dynamics functors between graphs and parameterized dynamical systems
- Vertical composition is by composition in FinSet and in Para(Dynam)
- Horizontal composition and monoidal products are by pushouts and coproducts in Para(Dynam)
- Finite sets in the feet of the cospans interpreted as linearly parameterized dynamical systems with no parameter variables and identically zero vector fields
- Symmetric monoidal double category Open(Para(Dynam + )) of open conically parameterized nonnegative dynamical systems
- Projection functor π S : Para(Dynam) → FinSet
- Left adjoint Z : FinSet → Para(Dynam)
- Symmetric monoidal double category of Z-structured cospans
- Double functor between open graphs and open parameterized dynamical systems
- Coproduct of the parameter variables for identified vertices
- Lax double functor LV : Open(FinGraph) → Open(Para(Dynam))
- Comparison cells defined using the morphisms of linearly parameterized dynamical systems
- Natural transformation α S : Z(S) → LV(Disc S)
Conclusion
- Regulatory networks are a tool to describe interactions between molecules in biochemical systems
- We studied regulatory networks, reaction networks, and parameterized dynamical systems
- We used signed graphs, Petri nets with signed links, and Lotka-Volterra dynamics
- We aimed to systematize the language and methods of describing, composing, and transforming scientific models
- We studied four different motifs in regulatory networks
- We discussed feedback loop analysis in system dynamics
- We studied open signed graphs and open parameterized dynamical systems
- We constructed a lax double functor between open signed graphs and open parameterized dynamical systems