In the middle of the 20th century, biology looked nearly finished: Darwin explained adaptation, genetics explained inheritance, and DNA had just been shown to carry the genes. The British biologist Conrad Hal Waddington was unconvinced. Nobody could say how genes reliably build all the tissues of a body, in the right places, at the right time. So he drew a picture.
Waddington imagined development as a rolling landscape cleft by branching valleys. A ball at the top represented a population of pluripotent cells; as it descended, each fork forced a choice, until the ball settled into one of a limited number of basins — skin, muscle, nerve. Cells stayed in their fates because they were "canalized", channelled by the valley walls. Beneath the sheet, he sketched a network of pegs and ropes: genes tugging the terrain into shape.
It was a brilliant metaphor with no mathematics behind it. "Without mathematical content," said James Briscoe of the Francis Crick Institute, the image "could not distinguish between alternative mechanisms, make quantitative predictions, or be falsified" — and it "has sometimes become a cliché rather than a meaningful explanation".
That is changing. Single-cell RNA sequencing now measures the activity of thousands of genes in thousands of individual cells at once, and the results can be projected into low-dimensional maps — UMAPs — where cells of the same type cluster together and lineages trace paths through time. Mapped as a network flow, the data looks strikingly like Waddington's landscape: cells funnelled down valleys into basins that correspond to distinct cell types.
In 2007 the biologist Sui Huang showed that blood progenitors committing to red or white blood cells form a real bifurcation, governed mainly by two proteins, GATA1 and PU.1. Push up on a pool of water suspended in a sheet and it spills sideways into lower pools: these basins are attractors, states the cells are inevitably drawn towards. Research groups including Briscoe's and mathematician David Rand's have since described many fate decisions the same way, and found that a small number of key genes drive each decision while the rest of the network merely enacts it — a structure that evolution would favour, because deep, well-separated attractors make development robust to molecular noise.
The updated picture complicates the original. Height in the landscape reflects the stability of a cell state, and "time" is really pseudo-time, an ordering of cells by how far their gene-expression profiles have changed — which works while cells cross unstable ridges but breaks down inside a basin, where profiles meander. Most importantly, the landscape is not fixed. "In our framework, the landscape is a feature of each cell and is not fixed," Rand said. "Attractors appear and disappear through bifurcations, and the routes connecting them change." There is no universal gravity; cells build and reshape the terrain they navigate by signalling to one another and tugging on their neighbours as tissues fold and grow.
That reframing carries practical weight. If you know where the bifurcations lie, stem-cell engineers can in principle steer cell populations towards a desired state instead of relying on trial and error. Huang is now applying the same lens to cancer — not as a malfunctioning cell but as one more attractor that many routes can reach. The therapeutic question shifts from "how do we kill them all" to "can we drive them somewhere else", pushing tumour cells back into a benign basin where they stop dividing or die. "The landscape framework provides a rational basis" for that kind of design, Briscoe said: not just a metaphor, but an atlas of terrain that is only now being charted.
