Gene regulatory networks
A gene regulatory network describes several genes or gene products that influence one another. One regulator may activate another gene, repress it, or participate in a feedback loop.
From one gene to a network
For one regulated product \(x\), we might write
\[\frac{dx}{dt}=\text{regulated production}-\text{degradation}.\]For many products,
\[\boxed{\frac{dx_i}{dt}=F_i(x_1,x_2,\ldots,x_n)-\delta_i x_i.}\]The function \(F_i\) contains all regulatory inputs controlling production of component \(i\).
How to read a regulatory diagram
An activation arrow means increasing \(A\) tends to increase production of \(B\). A repression symbol means increasing \(A\) tends to reduce production of \(B\).
The diagram tells us the structure of the network. The equations tell us how strongly and how quickly those interactions operate.
A simple activation cascade
If \(A\) is an external input, one possible model is
\[\frac{dB}{dt}=\alpha_B\frac{A^n}{K_B^n+A^n}-\delta_BB,\]\[\frac{dC}{dt}=\alpha_C\frac{B^m}{K_C^m+B^m}-\delta_CC.\]First \(A\) controls \(B\). Then the changing level of \(B\) controls \(C\). The signal therefore propagates through the network.
Why downstream responses are delayed
After \(A\) changes, \(B\) cannot jump instantly to its new level because it must be produced and degraded through time. Since \(C\) responds to \(B\), it begins changing later.
Thus a regulatory cascade can create a delay even when no explicit delay term appears in the equations.
A model-generated cascade
The graph below solves a two-stage activation cascade after the input \(A\) is switched on at \(t=0\). Parameters are chosen so that both regulated levels lie between 0 and 1.
Negative regulation
If \(A\) represses \(B\), a common production function is
\[F_B(A)=\alpha_B\frac{K^n}{K^n+A^n}.\]Then
\[\frac{dB}{dt}=\alpha_B\frac{K^n}{K^n+A^n}-\delta_BB.\]Increasing \(A\) lowers production of \(B\).
Feedback changes the behaviour
A feedback loop occurs when a component influences something upstream of itself. Feedback is one of the main reasons regulatory networks can behave very differently from isolated genes.
Negative feedback
Here \(A\) promotes \(B\), but \(B\) suppresses \(A\). If \(A\) becomes too large, the resulting increase in \(B\) pushes \(A\) back down.
Negative feedback can stabilise expression, reduce sensitivity to disturbances and, with suitable nonlinearities or delays, contribute to oscillations.
Positive feedback
A product that promotes its own production creates positive feedback. A simple model is
\[\boxed{\frac{dA}{dt}=b+\alpha\frac{A^n}{K^n+A^n}-\delta A.}\]The constant \(b\) represents basal production.
If \(A\) rises, activation becomes stronger, which can make \(A\) rise further. Strong nonlinear positive feedback can create more than one stable expression state.
Production and degradation curves
For a one-variable feedback model, an equilibrium occurs wherever
\[\text{production}(A)=\text{degradation}(A).\]Graphing these two quantities is often the clearest way to understand why multiple equilibria can occur.
Bistability: a biological switch
Suppose the system has a low stable equilibrium and a high stable equilibrium. Then the same regulatory network can remain in either of two persistent expression states.
A sufficiently strong perturbation can move the system across the unstable threshold from one state to the other. This is called bistability.
The toggle switch
A classic two-gene network has mutual repression:
A simple model is
\[\frac{dA}{dt}=\frac{\alpha_A}{1+(B/K_B)^n}-\delta_AA,\]\[\frac{dB}{dt}=\frac{\alpha_B}{1+(A/K_A)^m}-\delta_BB.\]If \(A\) is high, it suppresses \(B\). Low \(B\) then removes repression of \(A\), reinforcing the high-\(A\) state. The reverse can also occur.
Nullclines reveal possible states
For two variables, the \(A\)-nullcline is where
\[\frac{dA}{dt}=0,\]and the \(B\)-nullcline is where
\[\frac{dB}{dt}=0.\]Their intersections are equilibria of the complete network.
The number and arrangement of these intersections can reveal whether the system has one state, multiple states or thresholds between them.
Oscillatory networks
Feedback does not always produce a steady state. Negative feedback combined with sufficiently strong nonlinearity and effective delay can generate repeated rises and falls in gene-product levels.
Biological examples include regulatory mechanisms associated with circadian rhythms and cell-cycle control.
A three-component negative-feedback loop
A conceptual loop can be written
A change in \(A\) takes time to pass through \(B\) and \(C\). By the time repression returns to \(A\), the system may have overshot. Repeated delayed correction can support oscillation under suitable parameter values.
Network motifs
Large regulatory networks contain small recurring patterns called network motifs. Common motifs include autoregulation, feed-forward loops, mutual repression and feedback cycles.
The same motif can appear in different organisms because particular structures are useful for particular dynamical tasks.
Feed-forward loops
In a feed-forward loop, \(A\) regulates \(B\), and both \(A\) and \(B\) regulate \(C\):
Depending on the signs and logic of the interactions, such a motif can filter short signals, create delays or accelerate responses.
How multiple regulators are combined
A gene may receive several regulatory inputs. The production function must specify how they combine.
For example, if two activators are both required, a simple phenomenological model may multiply their activation functions:
\[F(A,B)=\alpha\frac{A^n}{K_A^n+A^n}\frac{B^m}{K_B^m+B^m}.\]Other biological mechanisms may require additive, competitive or mechanistic binding models instead.
Network structure versus parameter values
The arrows alone do not determine the behaviour. The same network structure can behave differently when production rates, degradation rates, thresholds or Hill coefficients change.
Mathematical analysis therefore needs both the topology of the network and its parameters.
Stability
For a system
\[\frac{dx}{dt}=F(x),\]we first find equilibria and then examine whether nearby perturbations return to them or move away.
For multidimensional gene networks this is commonly studied using the Jacobian matrix and its eigenvalues.
Bifurcations
Changing a regulatory parameter can alter the number or stability of equilibria. For example, increasing positive-feedback strength can transform a system with one stable state into a bistable switch.
Such qualitative changes are bifurcations and are important for understanding transitions between cellular states.
Deterministic versus stochastic networks
ODE models describe smooth concentration dynamics. Real gene regulation occurs through discrete molecular events: promoter switching, transcription, translation and degradation.
When copy numbers are small, random fluctuations can move cells between states or make genetically identical cells behave differently.
Stochastic gene regulatory networks can therefore explain variability that deterministic trajectories alone cannot represent.
Why gene regulatory networks matter
They provide mathematical models for cell differentiation, developmental decisions, stress responses, circadian regulation, synthetic biological circuits, signalling responses and many other cellular processes.
A practical modelling workflow
Identify the important genes or proteins. Draw activation and repression arrows. Convert each regulatory input into a production function. Add degradation terms. Write the coupled ODEs. Find equilibria and nullclines. Test stability and parameter sensitivity. Look for feedback, bistability, oscillations or bifurcations. Add stochasticity when molecular fluctuations are biologically important.