Day 45 · 2026.08.06

Mathematics & Biology

Form, ancestry, and folding — the equations hiding inside living systems
"The form of an object is a diagram of forces." — D'Arcy Thompson, On Growth and Form

Population Dynamics

Why ecosystems never settle at equilibrium
Dynamical Systems
Intuition

More rabbits, well-fed foxes; more foxes, the rabbits are eaten out; fewer rabbits, the foxes starve; fewer foxes, the rabbits recover. The loop never settles at the equilibrium point, because every link in it lags: how many foxes there are now is set by the rabbits they ate earlier. Negative feedback with delay does not produce rest, it produces oscillation.

prey x predator y equilibrium
Phase plane: each closed orbit is one boom-and-bust cycle, its amplitude fixed entirely by the initial condition.
Formal definition
$$\dot{x} = \alpha x - \beta xy, \qquad \dot{y} = \delta xy - \gamma y$$

$x,y$ are prey and predator densities; $\alpha$ is the prey's intrinsic growth rate, $\gamma$ the predator's death rate without food, $\delta$ the efficiency of turning eaten prey into offspring. The term $\beta xy$ is the encounter rate — assume the two mix at random and encounters go as the product of densities, which is exactly chemistry's law of mass action.

Why it is beautiful

The system hides a conserved quantity $V = \delta x - \gamma\ln x + \beta y - \alpha\ln y$, constant along every orbit, acting just like energy in mechanics — so all orbits are closed curves. That neutral stability also means structural instability: add a touch of density limitation on the prey and the closed orbits immediately collapse into spirals. A model too clean to be beautiful is usually a model too idealized to be true.

And the same biological assumption written one generation at a time, $x_{n+1}=rx_n(1-x_n)$, period-doubles into chaos as $r$ grows. Discrete or continuous is not a modelling detail; it decides which class of system you have.

Applications
  • Volterra's principle: killing indiscriminately (a pesticide that takes out pest and natural enemy alike) on average increases the prey and reduces the predator — which is why broad-spectrum pesticides so often trigger secondary pest outbreaks.
  • Same family of equations: the SIR threshold theorem $R_0>1$ is the same bifurcation analysis; the replicator equation is equivalent to Lotka–Volterra, so competing species and competing strategies share one dynamics.
In one line · a question
Populations do not converge to equilibrium because feedback takes time; delayed negative feedback is the universal recipe for oscillation.
Consider: adaptive rate limiting is delayed negative feedback too. Ecology took a century to learn that culling predators alone will not save the prey — what is the corresponding lesson for control systems?

Turing Patterns

How the force that erases differences manufactures pattern
PDE · Morphogenesis
Intuition

Diffusion is the force that erases differences: ink in water always ends up uniform. In 1952 Turing proved the opposite — when two mutually reacting substances diffuse at the same time and at different speeds, the uniform state spontaneously collapses into spots or stripes.

The mechanism fits in a sentence: the activator promotes itself and also the inhibitor, the inhibitor suppresses the activator, and the inhibitor diffuses much faster. Any random bump is amplified where it stands while broadcasting inhibition far afield, suppressing bumps nearby — local activation plus long-range inhibition, so peaks can only appear a fixed distance apart.

local activation long-range inhibition net influence vs distance wavelength λ the periodic solution left standing
A net-influence curve that is positive nearby and negative far off picks out one wavenumber, and the uniform state collapses into stripes.
Formal definition
$$\partial_t u = f(u,v) + D_u\nabla^2 u, \qquad \partial_t v = g(u,v) + D_v\nabla^2 v$$

$u,v$ are activator and inhibitor concentrations, $f,g$ the local reactions, $D$ the diffusion coefficients. The operator $\nabla^2$ measures how far a point sits below its neighbourhood average — positive means the point is a "dip" that matter will flow into, which is precisely what diffusion does.

Turing instability demands a seemingly contradictory pair of conditions: the uniform steady state is stable without diffusion, yet once diffusion is switched on some nonzero wavenumber $k$ has a positive growth rate. This requires $D_v \gg D_u$, and the selected $k^*$ corresponds to a wavelength $\lambda \sim \sqrt{D/\text{reaction rate}}$.

Why it is beautiful

Two individually stable mechanisms couple into an instability, and the term responsible for the instability is the one whose whole job is smoothing — hence the name diffusion-driven instability. Lovelier still is where the scale comes from: the wavelength depends only on local parameters, not on the size of the container. It is an intrinsic length carried by the medium rather than something the boundary drew, which is why a leopard's spots and a house cat's are about the same size. Form needs no blueprint, only a set of rules that destabilize on their own.

Applications
  • Zebrafish stripes are an experimentally confirmed Turing system: Shigeru Kondo ablated pigment cells with a laser, and the stripes that regrew followed the reaction–diffusion prediction rather than simply restoring the original. Digit number is set by a BMP/Sox9/Wnt Turing network (Sharpe lab, 2014) — tune the parameters down as the model says and mice grow more, thinner digits.
  • Non-living versions: dune spacing, banded vegetation in drylands, the BZ reaction; the computational echo is neural cellular automata.
In one line · a question
A pattern need not be drawn; it can be what is left over after a uniform state loses stability.
Consider: if "what grows where" is written in no coordinate table anywhere in the genome but emerges from the dynamics, where should the boundary of "genetic information" be drawn?

Phylogenetic Trees

Inferring yesterday's branchings from today's leaves
Combinatorics · Statistics
Intuition

All you hold are sequences from species alive today (or viruses sampled today), and you must reconstruct the family tree: who split from whom last, and how long ago. The ancestors left no samples, so everything is inferred from how alike the cousins are. There are two difficulties: the number of candidate trees explodes, and similar does not mean related — two independently evolving lineages can drift into resemblance by accident.

A B C D true tree (A,B),(C,D) A C B D parsimony infers (A,C),(B,D)
Long-branch attraction: A and C each evolve fast (thick branches), and their coincidentally identical bases are misread as evidence of common descent.
Formal definition

There are $(2n-5)!!$ unrooted binary trees on $n$ leaves: at $n=20$ that is already about $2\times 10^{20}$, so enumeration is hopeless.

Maximum likelihood models base substitution as a continuous-time Markov chain: with rate matrix $Q$, the transition probability along a branch of length $t$ is $P(t)=e^{Qt}$, where $t$ is measured in expected substitutions, not years. The likelihood at each site sums over all ancestral states — naively $4^{\text{internal nodes}}$ terms — but Felsenstein's pruning algorithm folds the tree bottom-up, storing only a length-4 conditional likelihood vector per node, so one post-order traversal suffices. This is the same algorithm as belief propagation on a tree in graphical models.

Why it is beautiful

The four-point condition: a distance matrix $d$ is exactly realizable by some weighted tree if and only if, for any four points, the largest two of the three pairings $d_{ij}+d_{kl}$, $d_{ik}+d_{jl}$, $d_{il}+d_{jk}$ are equal. A purely combinatorial object — a tree — turns out to be completely characterized by inequalities on continuous distances. That is what a tree metric is.

The converse is equally deep: maximum parsimony is statistically inconsistent when two lineages evolve unusually fast and are not neighbours on the true tree — more data makes it more certain of a wrong tree. An estimator can be intuitively unimpeachable and computationally efficient while converging, in the limit, to the wrong answer; consistency has to be proved, not felt.

Applications
  • Viral tracing: SARS-CoV-2 lineage assignment and molecular-clock dating are real-time maximum-likelihood tree building; Omicron appeared as a startlingly long branch, and that branch is what suggested the hypothesis of prolonged evolution inside a chronically infected host.
  • Elsewhere: treating somatic mutations from different biopsy sites as "species" reconstructs a tumour's clonal tree; the same reconstruction for HIV transmission chains has been accepted as courtroom evidence; language phylogenies and manuscript stemmatics run the identical algorithm — all it takes is vertical inheritance plus random change.
In one line · a question
Tree building is hard not because it is slow but because similarity lies; the model exists to subtract "identical by chance" from "identical by descent."
Consider: long-branch attraction arises because parsimony quietly assumes substitutions are rare. Which estimator you use daily hides an assumption you have never tested?

Protein Folding

Difficulty is a property of the landscape, not of the dimension
Optimization · Geometry
Intuition

An amino acid chain is synthesized and within milliseconds curls itself into a precise three-dimensional shape — the same one every time. Levinthal did the arithmetic in 1969: 100 residues, a conservative 3 conformations each, gives $3^{100}\approx 10^{47}$ possibilities; trying one every $10^{-13}$ seconds would still take $10^{27}$ years. There is only one conclusion: folding is not searching, it is going downhill.

Picture conformation space as terrain whose height is free energy. If it were a golf course — mostly flat, one hole somewhere — blind search really would be hopeless; if it is a funnel, sloping from all sides toward one bottom, then sliding down from any starting point gets you there.

free energy golf-course landscape: blind search over 10⁴⁷ states native state folding funnel: any start slides to the bottom
Those conformations all still exist; the difference is that they are never visited.
Formal definition

Energy landscape theory: define a free energy $F(x)$ over configuration space, with the native state as the global minimum. The key quantity is frustration — in a random sequence the interactions conflict (optimize one contact and another suffers), so the landscape is rugged and full of traps; sequences selected by evolution instead satisfy the principle of minimal frustration, where nearly every contact in the native state is close to its own optimum at once, and the landscape as a whole becomes a funnel.

Why it is beautiful

The resolution of Levinthal's paradox is not a faster search but the observation that difficulty is a property of the landscape, not of the dimension. Those $10^{47}$ conformations are still there; they are simply never visited.

The insight reappears word for word in AI: a deep network's loss function is high-dimensional and non-convex, which classical optimization theory says should be hopeless, yet gradient descent trains it anyway — because an overparameterized loss landscape is also funnel-like, dominated by saddle points rather than local minima. "High-dimensional and non-convex" was never sufficient for hard; the geometry of the landscape is what matters — and geometry can be chosen. Biology chooses it by evolution, we choose it by architecture and initialization.

Applications

AlphaFold2's core move is not simulating physics but reading co-evolutionary signal: if two residues touch in the folded structure, a mutation in one usually needs a compensating mutation in the other, so across a multiple sequence alignment those two columns are correlated. The trouble is that correlation propagates ($i$–$j$ plus $j$–$k$ fabricates an apparent $i$–$k$), and the fix is to fit a Potts model and invert its coupling matrix — precisely the standard move from "correlated" to "conditionally independent" in graphical models, which AlphaFold replaces with learnable attention. Methods like RFdiffusion go further, running diffusion models over structure space for de novo design, generating enzymes that do not exist in nature.

In one line · a question
Folding works not because the search is fast but because evolution carved the landscape into a funnel; the way out of a hard problem is often to reshape the terrain rather than to search it faster.
Consider: AlphaFold returns a static structure, while function usually depends on a conformational ensemble and allosteric switching — what still lies between "predicting structure" and "understanding folding"?

Going deeper

How well established are Turing patterns in actual biology?
Less than the popular accounts claim, more than the sceptics allow. A good share of the canonical examples — leopard rosettes, butterfly wing spots — are in fact set by cell migration or pre-existing morphogen gradients, not by genuine reaction–diffusion. The established cases all come with an identified molecular pair: zebrafish stripes, digit number, hair-follicle spacing — and zebrafish in fact relies on contact-dependent interactions between pigment cells, a cellular rather than chemical version of local activation with long-range inhibition. The safer statement is that the Turing mechanism is a solution repeatedly and independently invented, not the universal language of development — it turns up everywhere because it is the minimal way to generate discrete structure in a uniform medium.
Can evolution be written as an optimization problem?
Only locally. Fisher's fundamental theorem says the rate of increase in mean fitness equals its additive genetic variance, which reads like gradient ascent — provided the landscape is fixed. But real fitness depends on everyone else's strategy: when predators get faster, the payoff to "running fast" changes. The landscape deforms along with the population distribution, and optimizing a fixed objective becomes game dynamics. Hence the Red Queen: every species keeps improving while mean fitness does not rise. This is the same mathematical difficulty wearing different faces in the non-stationarity of multi-agent RL and the cyclic oscillations of GAN training.
Must life be described with continuous mathematics?
No — and in the cases that matter, it must not be. Differential equations assume concentrations are continuous, which holds only when molecule counts are large. A transcription factor often exists in a few dozen copies per cell, where "half a molecule" is meaningless and fluctuations are the same size as the mean; the right object is the chemical master equation and the practical algorithm is Gillespie stochastic simulation. Nor is this a small correction to precision: noise produces qualitatively new behaviour, such as random flipping of genetic switches and genetically identical cells taking different fates (which is how bacterial persisters survive antibiotics). What the continuous approximation discards is not a few decimal places but an entire class of phenomena.