The Inescapable Equilibrium: What Math Tells Us About Treating Alzheimer's
In the mathematics of disease, some equilibria cannot be escaped.
Imagine standing at the edge of a vast, invisible valley — a landscape shaped not by geography, but by the chemistry of your own brain. No matter where you start, no matter which path you take, the terrain pulls you inexorably toward the same bottom. You can climb the walls. You can scramble sideways. But the valley keeps dragging you back.
This is the picture that emerges from a new mathematical analysis of Alzheimer's disease, published by researchers Ruoyun Lang and Hui Zhou from the University of Science and Technology of China and Hefei Normal University. Their work, which rigorously proves a fundamental property of the disease's underlying biology, carries an uncomfortable message: the human brain, once caught in the dynamics of Alzheimer's pathology, tends always — always — toward a固定的 disease state. No matter where the system starts. No matter how promising the early interventions look. The math, it turns out, is unforgiving.
But here's where the story takes an unexpected turn. This mathematical inevitability is not a death sentence. It's a map. And maps, as any explorer knows, are most valuable not when they show you where you can't go, but when they reveal the paths you haven't tried yet.
The Science
To understand what Lang and Zhou accomplished, you need to understand what they're modeling. Alzheimer's disease — the leading cause of dementia among elderly populations worldwide, now affecting an estimated 50 million people and projected to triple by 2050 as populations age — is increasingly understood as a disorder of aggregation. Specifically, it's a disorder of a protein fragment called β-amyloid, or Aβ.
Aβ is produced continuously in the brain through the routine cleavage of a larger protein called amyloid precursor protein. Normally, this production and clearance would be in balance, like water flowing through a bathtub with the drain open. But in Alzheimer's, something goes wrong. The Aβ fragments, each just 36-43 amino acids long, begin to clump together. First into soluble clusters called oligomers — sometimes called "soluble toxins" in the literature because they're increasingly recognized as the most damaging form. Then, into larger and larger aggregates, eventually forming the dense, insoluble plaques that dot the brains of Alzheimer's patients on autopsy.
But Aβ doesn't act alone. In 2009, a research team led by Lauren and colleagues made a striking discovery: Aβ oligomers bind to a specific receptor on neuronal surfaces called cellular prion protein, or PrP^C. This binding creates a neurotoxic complex — an Aβ-PrP^C pairing — that disrupts synaptic function and drives cognitive decline. PrP^C, it turns out, acts as a critical mediator, essentially serving as the doorway through which Aβ oligomers inflict their damage on the brain's communication networks.
This discovery opened a new frontier in understanding Alzheimer's. It also created a new target for mathematical modeling. Because if Aβ oligomers, PrP^C receptors, and their toxic complex are all interacting — each affecting the others, each being produced and degraded at particular rates — then the system has dynamics. It evolves. And dynamics, as any physicist or ecologist knows, can be described with mathematics.
The model at the center of Lang and Zhou's analysis traces its origins to work by Helal, Docter, and colleagues published in 2014. That earlier research had proposed a system of differential equations describing how Aβ plaques grow, how Aβ oligomers accumulate, how PrP^C receptors are expressed, and how the toxic Aβ-PrP^C complex forms and breaks apart. The model took the form of four coupled equations tracking:
The concentration of Aβ plaques, $A(t)$, which grow when new plaques form from oligomers and shrink when plaques are degraded; The concentration of soluble Aβ oligomers, $u(t)$, which are produced continuously, cleared by various mechanisms, and converted into plaques; The concentration of cellular prion protein, $p(t)$, which is also produced and degraded, and which binds to oligomers to form the toxic complex; And the concentration of the Aβ-PrP^C complex itself, $b(t)$, which forms when oligomers bind to prion protein and dissociates over time.
The equations that describe these four variables involve a dozen or so parameters — production rates, degradation rates, binding and unbinding rates — each representing some real biological process. The system is nonlinear, meaning the equations contain terms where variables are multiplied together, which makes them dramatically harder to solve than the linear equations you might remember from high school algebra. Nonlinear systems can behave in surprising ways: they can oscillate, they can bifurcate into multiple regimes, they can exhibit chaotic behavior where tiny changes in initial conditions produce wildly different outcomes.
What Helal and colleagues had shown was that this system has a unique equilibrium — a single point where all four variables settle into a stable, unchanging state. They proved this equilibrium always exists and is unique, and they demonstrated that it's locally asymptotically stable, meaning that if you start very close to this equilibrium, you'll converge toward it over time. This equilibrium, crucially, corresponds to the pathological state of Alzheimer's disease: persistent plaques, oligomers, prion protein, and toxic complexes all present, all damaging neurons.
But local stability is a limited concept. It only tells you what happens near the equilibrium. What Helal's team couldn't prove — what had remained an open question for over a decade — was whether the equilibrium is globally asymptotically stable. That is: no matter where you start in the vast landscape of possible brain states, no matter what the initial concentrations of plaques, oligomers, prion protein, and complexes happen to be, will the system always, inevitably, converge to this pathological equilibrium?
In pure mathematics, this is a genuinely difficult question. The system is four-dimensional, meaning it lives in a space with four coordinates (one for each variable), and four-dimensional nonlinear systems don't come with easy proofs. To make the problem tractable, Helal and colleagues had to impose a specific constraint on the parameters — a mathematical condition that, while technically useful, had no clear biological meaning. They assumed that certain combinations of production and degradation rates had to fall within a narrow numerical range for their proof to work. The question was whether this constraint was a genuine requirement of the biology, or merely an artifact of the mathematical tools they had chosen.
Lang and Zhou set out to answer this question. Using a more sophisticated approach — drawing on the theory of competitive dynamical systems and a technique called the second compound matrix method, originally developed by mathematician James Muldowney in 1990 — they were able to prove the global asymptotic stability of the equilibrium without any artificial parameter constraints. Their Theorem A, as they call it, establishes mathematically that the unique equilibrium of the model is globally asymptotically stable. In the context of the early-stage disease, when new plaque formation is negligible, they prove even stronger results: the three-dimensional reduced system (tracking oligomers, prion protein, and their complex) always converges to its unique positive equilibrium, regardless of initial conditions.
This is the mathematical heart of the paper. The system is unconditional. It doesn't depend on parameters falling into special ranges. It doesn't require that production rates and degradation rates be finely tuned. The mathematics says, simply and powerfully: you end up at the disease equilibrium.
What They Found
The proof itself is a technical achievement, but its implications are best understood through the numerical simulations that Lang and Zhou ran to illustrate and validate their theoretical results. These simulations — which trace how the system evolves over time under different parameter values — serve as windows into the mathematical landscape the equations describe.
The first set of simulations examines how degradation rates affect the system's long-run equilibrium. In biology, degradation rates represent how quickly the body clears a substance — how fast plaques are removed, how quickly oligomers are broken down, how rapidly prion proteins are recycled. In the model, the parameter $\mu$ controls the degradation rate of Aβ plaques, $\gamma_u$ controls the degradation of soluble oligomers, and $\gamma_p$ controls the degradation of prion protein.
What the simulations show is striking. When $\mu$ is very low — meaning plaques are cleared very slowly — the equilibrium concentration of plaques is high. When $\mu$ is increased to a moderate value, plaques decline. And when $\mu$ is increased further, plaques decline even more. The relationship is clean and consistent: faster plaque clearance means lower equilibrium plaque levels.
Effect of Plaque Degradation Rate μ on Equilibrium Aβ Plaques
| Label | Value |
|---|---|
| Baseline (μ=0.3) | 100 |
| Moderate increase (μ=0.5) | 62 |
| High increase (μ=0.8) | 38 |
The same pattern holds for the oligomer degradation rate $\gamma_u$. Higher $\gamma_u$ means faster oligomer clearance, which translates directly into lower equilibrium oligomer concentrations. And crucially, lower oligomer concentrations mean less raw material available to form the toxic Aβ-PrP^C complex.
The effect of the prion protein degradation rate $\gamma_p$ is more nuanced, which reflects the dual role of PrP^C in the system. Prion protein isn't just a target for Aβ oligomers — it's also produced continuously in the brain and has its own production and degradation dynamics. Higher $\gamma_p$ means faster prion protein turnover, which initially might seem protective (less PrP^C available to bind oligomers). But the simulations show that the relationship is not straightforward: the effect of $\gamma_p$ depends on where you are in the parameter space and what other rates are doing.
Perhaps most revealing are the simulations examining the binding rate $\tau$ — the rate at which Aβ oligomers and PrP^C come together to form the toxic complex. In the baseline scenario, with $\tau = 1.0$, the system settles into an equilibrium where oligomers, prion protein, and complex are all present at measurable concentrations. When $\tau$ is reduced to 0.5 — meaning oligomers and prion protein bind less readily — the equilibrium concentration of the toxic complex drops significantly. Oligomer and prion protein concentrations rise slightly (because they're not being removed as efficiently through complex formation), but the overall toxic burden, as measured by complex concentration, is substantially reduced.
These simulations validate the theoretical finding of global stability in the most direct way possible: by showing that the system, starting from random initial conditions, always converges to the same equilibrium point. No matter where the simulation begins — whether it starts with high plaques and low oligomers, or low plaques and high oligomers, or any other combination — it ends up at the same destination. The valley, to return to our geographical metaphor, always pulls you to the same bottom.
Effect of Binding Rate τ on Toxic Aβ-PrP^C Complex
| Label | Value |
|---|---|
| Baseline (τ=1.0) | 100 |
| Reduced binding (τ=0.5) | 45 |
| Further reduced (τ=0.2) | 18 |
But here's the insight that transforms this from a mathematical curiosity into something with real clinical implications: the simulations also show that you can change the equilibrium. You can't escape it — the system will always find its way back to some stable state. But you can change what that stable state looks like. By manipulating the parameters — by making plaques degrade faster, or by reducing the binding affinity between oligomers and prion protein, or by increasing the clearance of oligomers — you can shift the equilibrium toward a state where toxic complex concentrations are lower, where oligomer levels are reduced, where the damage to neurons is minimized.
This is the therapeutic message buried in the mathematics. The disease equilibrium is inevitable. But the shape of that equilibrium is not fixed. And shaping it — reducing the toxic burden even if you can't eliminate it entirely — is what treatment might actually look like.
Why This Changes Things
The global stability result matters for several reasons, each of which reveals something important about how we should think about treating Alzheimer's disease.
First, it settles a long-standing mathematical question. For over a decade, the proof of global asymptotic stability for the Aβ-PrP^C model had relied on an artificial parameter constraint — a condition that didn't correspond to any known biological requirement. Mathematicians had wondered: is this constraint necessary because the biology demands it, or is it an artifact of the proof technique? Lang and Zhou's work provides a definitive answer: it's an artifact. The system is globally stable without any constraints on the parameters. The previous proof was limited by its tools, not by the underlying biology.
This might seem like a technical matter, interesting only to specialists. But it has profound implications for how we think about treatment. If the constraints were genuinely required — if the global stability only held when certain rates fell into special numerical ranges — then therapy would be impossibly constrained. You'd need to hit very specific parameter combinations to achieve the desired dynamics. But since the constraints were artificial, those constraints disappear. The system is stable across the full range of biologically plausible parameters. The therapeutic playing field is much larger than anyone realized.
Second, the result clarifies what "cure" means — and isn't — in the context of Alzheimer's. The mathematical inevitability of the disease equilibrium might sound like bad news: you're going to get sick, you're always going to be sick, no intervention can change that. But this interpretation misunderstands what the model is actually showing. The equilibrium isn't a binary state — healthy or diseased — it's a continuous spectrum. There are many possible equilibria, each with different concentrations of plaques, oligomers, prion protein, and toxic complex. The mathematical inevitability is that the system will find an equilibrium. What that equilibrium looks like is not fixed.
To use an analogy: consider a thermostat controlling room temperature. The thermostat always settles on some temperature — it doesn't oscillate wildly or behave chaotically. But the temperature it settles on depends on the setting you choose. Turn up the heat, and you get a different equilibrium than if you turn it down. Similarly, Alzheimer's disease will always settle into some pathological equilibrium, but the details of that equilibrium depend on the biological "settings" — the production rates, degradation rates, binding affinities that characterize an individual's physiology.
Effect of Oligomer Degradation Rate γ_u on Equilibrium Aβ Oligomers
| Label | Value |
|---|---|
| Baseline (γ_u=0.3) | 100 |
| Increased clearance (γ_u=0.6) | 52 |
| High clearance (γ_u=1.0) | 31 |
This reframing has practical consequences for how we design therapies. Current approaches to Alzheimer's treatment have often focused on reducing Aβ production or clearing Aβ plaques — attacking the upstream sources of the pathology. The monoclonal antibodies that have dominated recent clinical trials — drugs like lecanemab and donanemab — work by binding to Aβ and helping the immune system clear it. These treatments can reduce amyloid burden significantly, and the latest data suggests they can modestly slow cognitive decline in early-stage patients.
But the model suggests that targeting Aβ alone may be insufficient. If the system always returns to a disease equilibrium, then clearing plaques or reducing oligomers might be like bailing water out of a boat without fixing the hole. The equilibrium adjusts. Plaques decline, but if oligomer production and complex formation continue apace, the system simply settles into a new configuration — one with less plaque, perhaps, but still with persistent toxic complex.
The therapeutic strategy the model points toward is different: instead of attacking Aβ alone, you should target the interaction between Aβ and prion protein. The binding rate $\tau$ — the rate at which oligomers and PrP^C come together — turns out to be a powerful lever. Reducing $\tau$ doesn't just slow complex formation; it reshapes the entire equilibrium, reducing toxic complex concentration substantially while only modestly affecting oligomer and prion protein levels.
This is the insight that makes the clinical implications concrete. The model suggests that drugs targeting the Aβ-PrP^C interaction — perhaps by blocking the binding site on PrP^C, or by destabilizing the toxic complex directly — could achieve something that anti-amyloid antibodies alone cannot: sustained reduction of the most damaging species in the disease pathway.
Third, the result illuminates the importance of timing. Lang and Zhou's analysis focuses specifically on the early stages of Alzheimer's, when new plaque formation is negligible — what the paper calls "when the formation rate of new plaques is zero." This might seem like a limiting assumption, but it's biologically meaningful. In the earliest phases of the disease, before extensive plaque deposition, the system dynamics are governed by the three-variable subsystem tracking oligomers, prion protein, and complex. This subsystem also converges to a unique equilibrium that is globally asymptotically stable. The implication is that even early intervention — before plaques become extensive — faces the same fundamental challenge: the system wants to settle into a disease state.
But early intervention also has an advantage the model highlights: it's easier to shift an equilibrium when you're close to it. If you've only just begun to accumulate toxic complex, reducing the binding rate $\tau$ has a bigger proportional effect than it would if the complex were already at high levels. The model doesn't guarantee that early intervention will work, but it suggests that early intervention has better odds than intervention at later stages, when the system has settled more deeply into its pathological equilibrium.
What's Next
The mathematical proof is solid, the numerical simulations are convincing, and the biological interpretation is compelling. But several important questions remain open — questions that will determine whether this theoretical framework actually translates into clinical benefit.
The first open question is whether the model captures enough biology to be clinically useful. Mathematical models are always simplifications. The four-variable system Lang and Zhou analyze is a highly reduced representation of Alzheimer's pathophysiology, ignoring many factors that are known to be important: the role of tau protein and neurofibrillary tangles, the influence of neuroinflammation, the effects of vascular health, the complex genetics of risk factors like APOE4. Whether the core dynamics captured by the model — the aggregation of Aβ, the binding to prion protein, the formation of toxic complex — are sufficiently central that interventions targeting them will actually alter clinical outcomes remains to be demonstrated.
The second open question is how to actually manipulate the parameters the model identifies as therapeutic targets. Reducing the binding rate $\tau$ sounds appealing in theory, but how do you drug an interaction between two proteins in the brain? The blood-brain barrier makes drug delivery to the central nervous system notoriously difficult, and protein-protein interactions are notoriously hard to block with small molecules. Antibodies can be designed to target specific proteins, but their size often impedes brain penetration. Any therapeutic strategy emerging from this model will need to solve the drug delivery problem in addition to the target validation problem.
The third open question is how individual variation affects the model's predictions. The analysis treats parameters as constants — fixed numbers that describe a generic human brain. In reality, these parameters vary substantially across individuals: people differ in their rates of Aβ production, in their expression of prion protein, in their capacity to clear metabolic waste products. A therapy that shifts the equilibrium in one person might not work the same way in another. The model provides a theoretical framework, but applying it to individual patients will require understanding the range of individual parameter values and how they change with age, disease stage, and genetic background.
The fourth open question is how the full model — including active plaque formation, not just the early-stage reduced version — behaves over the longer term. Lang and Zhou prove global stability for the system when plaque formation is absent, which is appropriate for early disease. But Alzheimer's is a progressive condition, and over time, plaque accumulation becomes substantial. How does the presence of actively forming plaques alter the dynamics? Do the equilibria change? Are there bifurcations — points at which small changes in parameters produce qualitative shifts in behavior? These are questions the paper raises but doesn't fully answer, leaving them as directions for future research.
Despite these open questions, the framework Lang and Zhou provide is a significant advance. Mathematical models of disease have historically been underutilized in neurology, where the complexity of the brain has often seemed to defy quantitative approaches. But the success of mathematical oncology — where models of tumor growth and treatment response are now routinely used in clinical decision-making — suggests that the same approach could work in neurodegeneration. The brain's complexity doesn't make modeling impossible; it makes it essential. Without quantitative frameworks, we're navigating the therapeutic landscape without a map, trying treatment after treatment without a clear theory of why things should work.
The map this paper provides isn't complete. It doesn't tell you exactly which drug to give, at what dose, to which patient. But it does something arguably more valuable: it provides a theoretical foundation for understanding why Alzheimer's is so hard to treat, and where treatment might most effectively intervene. The disease equilibrium is inevitable. The shape of that equilibrium is not. And the key to reshaping it, the mathematics suggests, lies in understanding — and ultimately manipulating — the interaction between β-amyloid and cellular prion protein.
As the global population ages, Alzheimer's disease has become one of the defining public health challenges of the century. The numbers are stark: tens of millions affected now, hundreds of millions at risk, trillions of dollars in healthcare costs, and a human toll that defies quantification. Every insight into the disease's mechanisms, every theoretical framework for understanding its dynamics, represents a step — however small — toward the goal of meaningful treatment.
Lang and Zhou's mathematical analysis won't appear in neurology textbooks. It's too technical, too specialized, too far from the clinical frontier. But the ideas it contains — the inevitability of the disease equilibrium, the importance of the Aβ-PrP^C interaction, the possibility of shifting equilibria even when you can't escape them — these ideas could shape how the next generation of Alzheimer's therapies is designed. The mathematics of disease is unforgiving. But it's also, in its own way, illuminating. And sometimes, the light it casts points toward paths that would otherwise remain invisible.