Rethinking Disease Through Complexity Science

I have long been interested in understanding medical problems through complexity science.

Modern medicine, particularly molecular biology, tends to break disease down into progressively smaller scales, looking for key proteins, signaling pathways, or cellular subpopulations that might explain it.

Yet one question keeps returning: must we understand every microscopic mechanism before we can intervene effectively in a macroscopic system?

01. Two timescales of life

I tend to distinguish two timescales in the evolution of living systems.

The first is the long timescale of evolution in its conventional sense. Through heredity, variation, and natural selection, life has developed its present forms of organization and regulation over vast spans of time.

The second is the shorter timescale of individual dynamics: a multicellular organism developing from a single cell, forming tissues and organs, maintaining homeostasis, aging, and eventually dying.

For an individual, that second process can be viewed as a complex dynamical system.

Its state is never fixed. From embryonic development through adulthood and aging, the system keeps changing. Homeostasis does not mean that every variable remains constant; it means that a living system maintains a relatively stable mode of operation amid continuous exchanges of matter and energy and external disturbances.

From this perspective, disease can be understood as a transition during individual dynamics from a state we regard as normal into a less common state, or one less favorable to the function of the whole. There is no absolutely sharp physical boundary between normality and pathology.

02. Disease may be another attractor

A complex dynamical system can have multiple stable or metastable states.

In some diseases, a disturbance may push the system out of its original basin of attraction and into another attractor.

Pathology, then, need not mean a loss of balance. The system may have established another self-sustaining state.

After myocardial infarction, for example, the initial ischemic event may have ended, while ventricular dilation, abnormal wall stress, fibrosis, and further remodeling form an ongoing positive-feedback loop.

Eliminating the original ischemic cause does not necessarily return the system to its earlier state.

Strictly speaking, living systems are open, nonequilibrium systems. The balance described here is closer to a nonequilibrium steady state or a dynamical attractor than thermodynamic equilibrium. Nor does every disease have a stable attractor: some pathological processes are continuously deteriorating trajectories.

Still, this offers a different perspective. Disease can be a different dynamical state, or configuration, of the system as a whole.

03. Scale, coarse-graining, and RG

Scale is another question that has long interested me.

Life is multiscale. Proteins, cells, tissues, organs, and whole organisms have different effective degrees of freedom. In statistical physics, coarse-graining and the renormalization group (RG) provide a framework for understanding effective descriptions at different scales.

A system can contain extremely complex microscopic interactions while its macroscopic behavior is governed by only a few effective variables.

When analyzing ventricular mechanics, for example, we usually consider geometry, material properties, local stress, and pressure–volume relationships. We do not need a molecular-dynamics simulation of every myosin molecule.

These macroscopic variables arise from microscopic components, but that does not mean direct manipulation of microscopic variables is the only way to control them.

Medicine raises a similar question. We may identify hundreds or thousands of differentially expressed genes in diseased tissue. How many genuinely drive disease dynamics, and how many simply respond to the system’s pathological state? Finding a molecular change does not establish that it caused the disease, much less that modulating it will provide effective treatment.

That is why I am skeptical of some current molecular-mechanism research, especially work that attempts to establish an entire pathological causal chain from expression data, correlations, and a small number of knockdown or overexpression experiments.

This is not an argument against molecular mechanisms or mechanistic research. It is an objection to a reductionist—and sometimes self-deceiving—style of molecular biology that has gone far enough.

Strictly speaking, relevant and irrelevant variables in RG cannot simply be identified with effective and ineffective therapeutic targets. The underlying way of thinking about scale, however, makes sense to me.

04. Does disease have a dominant scale?

Following this line of thought, I have a conjecture that still needs testing:

Many diseases may have a relatively well-defined dominant scale, while many changes at other scales are manifestations of that state at different levels.

A tissue-scale structural problem, for example, can affect organ function; a cellular metabolic problem can gradually affect tissues and organs.

I tend to think many pathological effects accumulate from local, lower scales toward higher ones. But feedback in the opposite direction certainly exists: changes in organ geometry can alter tissue stress, which then changes cellular state through mechanotransduction.

This directionality cannot be reduced to one-way causation.

A more useful question may be: despite bidirectional coupling across scales, are the variables that determine whether a disease persists or can be reversed concentrated mainly at one scale?

Consider valvular regurgitation. A structural valve abnormality can cause chamber volume overload, followed by myocardial hypertrophy, fibrosis, and widespread molecular-expression changes. If the initial structural abnormality remains, correcting downstream molecular changes may not eliminate the regurgitation. Conversely, repairing the valve may gradually improve many secondary physiological and molecular abnormalities.

This does not deny causal effects from higher scales to lower scales. It suggests that, in this particular problem, valve structure may have greater causal control.

05. Disease origin is not the same as control scale

We should distinguish the scale at which a disease originated from its current dominant control scale.

The first asks how the disease began. The second asks which variable is most likely to improve it now that it has reached its present state.

I am more interested in the latter.

Traditional medicine emphasizes etiology: identifying and addressing the primary disease. This is effective in many situations. Yet in many others, the system does not return to normal even after its original cause is treated.

Again, consider myocardial infarction. Coronary occlusion may initiate the problem, but cell death, scar formation, and ventricular remodeling can shift it to tissue or organ scale. Restoring coronary flow does not necessarily return a dilated ventricle to its original geometry. Directly changing mechanical loading or local wall stress might instead alter subsequent remodeling dynamics.

At that point, the most effective scale of intervention differs from the disease’s scale of origin. The control scale is not fixed either: it may change as the disease progresses.

For the same disease at different stages, interventions at molecular, cellular, tissue, or organ scale may have very different effects.