The Waddington Landscape: A Cell's Fate as a Ball Rolling Downhill

A metaphor that became a model

In 1957 the developmental biologist Conrad Waddington sketched a now-famous picture: a marble at the top of a hill, rolling down through a landscape carved with branching valleys, each valley leading to a different possible fate. The marble represents a developing cell; the landscape represents the constraints imposed by its genes; and each valley it can roll into represents a distinct, stable cell type — a valley labeled "neuron," another "skin," another "blood cell." Once the marble has settled into a valley and rolled past a ridge behind it, going back becomes very difficult, which is exactly the behaviour real cells show: differentiation is normally a one-way process.

For decades this was treated as a helpful illustration rather than a real quantity you could compute. That changed as systems biology developed tools to derive an actual quasi-potential landscape directly from a gene regulatory network's dynamics, turning Waddington's cartoon into a genuine mathematical object with real height values, real valleys and real ridges determined by the specific genes involved.

Where the valleys come from: bistable switches

The simplest and most studied source of a valley-ridge-valley structure is mutual inhibition between two master regulator genes: gene A represses gene B, and gene B represses gene A. This circuit has two stable steady states — high A/low B, or low A/high B — separated by an unstable state in between where neither wins. Plotted as a potential-energy-like surface, that unstable middle state is exactly a ridge, and the two stable states are the valley floors on either side of it. A cell expressing intermediate, roughly equal levels of both genes sits balanced on the ridge; the smallest push toward one side or the other commits it to a valley, after which the mutual-inhibition feedback actively reinforces the choice and makes reversal difficult. Chain many such bistable switches together, activating in sequence as development proceeds, and you get the branching, multi-level valley structure Waddington drew by intuition — a stem cell valley splitting into two, each of those splitting again, eventually reaching dozens of distinct terminal cell-type valleys.

Langevin dynamics: adding noise to the descent

Gene expression is not deterministic. Transcription happens in stochastic bursts, and in a single cell the number of molecules of a given regulatory protein can be small enough that random fluctuations are a significant fraction of the total. A model that captures this treats each cell's position on the landscape as following Langevin dynamics — deterministic motion down the local gradient, plus a random noise term at every step:

dx = -grad(U(x)) * dt + sqrt(2D) * dW

x      cell's position (a point in gene-expression space)
U(x)   the epigenetic landscape's height at x
D      noise intensity, reflecting stochastic gene expression
dW     a random increment (Brownian motion)

The gradient term pulls the cell downhill toward the nearest valley, exactly as Waddington's marble would fall under gravity. The noise term is what makes the model biologically honest: it is why two genetically identical cells, starting from the same point, can end up settling into different valleys, reproducing the cell-to-cell variability seen in real differentiating populations even in a perfectly controlled, genetically uniform culture. Occasionally, noise is also large enough to carry a cell back over a low ridge — a rare event under normal conditions, but exactly the kind of transition that induced pluripotency techniques (reprogramming an adult cell back toward a stem-like state) are, in this language, deliberately engineered to make far more likely.

What the landscape explains that a simple lineage tree does not

A textbook lineage tree tells you which cell types descend from which, but it says nothing about stability, reversibility, or why some transitions are common and others vanishingly rare. The landscape picture answers all three at once: valley depth corresponds to how stable a cell type is against perturbation, ridge height between two valleys corresponds to how hard it is to convert one type into the other, and the overall topology — how many times the landscape branches, and where — corresponds directly to the wiring of the underlying gene regulatory network. This is why the Waddington landscape, despite starting life as a hand-drawn cartoon, remains the standard conceptual and increasingly computational framework cell biologists reach for when discussing differentiation, reprogramming and the stability of cell identity.

Frequently asked questions

Try it live

Everything above runs in your browser — open Stem Cell Differentiation and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.