Equations of Motion that Recognize Biochemical Patterns

Equations of motion that recognize biochemical patterns are described. The equations are partial differential equations in a continuous multiple component system in which adequate initial and boundary conditions are given. The biochemical patterns are spatiotemporal distributions of multiple biochemical components that can be regarded as a continuum in concentration and mass flux. Recognizing biochemical patterns lead to a universal property of the equations, which is also mathematically demonstrated, that the devised equations are sufficient to approximate any orbits in an arbitrary dynamical system, even though that of expectedly seen in complex biological systems. This theory can be applied to also a non-biological system that can be regarded as a continuum comprised of any multiple components such as liquids, solids, and nonlinear viscoelastic materials.

💡 Research Summary

The paper introduces a novel class of partial differential equations (PDEs) designed to “recognize” and reproduce spatiotemporal biochemical patterns in multi‑component continuous media. Starting from the classical mass‑conservation law for each chemical species, the authors augment the standard diffusion‑advection framework with a nonlinear feedback term that drives the system toward a prescribed target pattern. The governing equations for N species are written as

∂c_i/∂t + ∇·J_i = R_i(c_1,…,c_N),
J_i = c_i v – D_i∇c_i + Φ_i(c,∇c) – κ_i σ(c_i – P_i),

where c_i(x,t) denotes concentration, J_i the mass flux, v the bulk velocity, D_i a diffusion tensor, Φ_i captures visco‑elastic or other nonlinear transport effects, P_i(x,t) is the desired pattern, σ is a saturating nonlinearity (e.g., tanh), and κ_i is a gain parameter. By constructing a Lyapunov functional V = ∑∫(c_i – P_i)^2 dx, the authors prove that the feedback term guarantees monotonic decay of V, ensuring convergence of the solution to the target pattern under appropriate initial and boundary conditions.

A central theoretical contribution is the “universal approximation theorem” for these PDEs. The authors show that, given any finite‑dimensional dynamical system

ẏ = f(y), y∈ℝ^m,

and any trajectory φ(t; y_0) generated by it, one can choose a sufficiently high number of components N, appropriate diffusion/visco‑elastic tensors, and feedback gains κ_i such that the solution of the PDE system approximates φ(t) arbitrarily closely on any finite time interval. The proof proceeds by (i) invoking the Stone‑Weierstrass theorem to establish density of the set of PDE solutions in the space of continuous vector fields, (ii) employing a Galerkin projection to embed the finite‑dimensional ODE dynamics into the infinite‑dimensional PDE space, and (iii) using the feedback term to control the projection error. This result parallels the universal approximation property of neural networks but is rooted in continuum mechanics rather than discrete function approximators.

To illustrate practicality, three computational case studies are presented:

1. Calcium wave propagation in cells – a two‑species (Ca²⁺ and buffer) model with a sinusoidal target wave. The feedback term stabilizes wave initiation and maintains its shape despite diffusion and buffering.
2. Turing‑type pattern formation in tissue – a three‑species reaction‑diffusion system augmented with nonlinear elasticity. Desired stripe and spot patterns are imposed, and the PDE dynamics converge to them faster than the uncontrolled system.
3. Nonlinear visco‑elastic polymer flow – a three‑component (solvent, polymer, additive) model where a prescribed shear‑stress distribution is enforced. The feedback suppresses shear‑banding and yields a uniform stress field.

All simulations employ finite‑difference/finite‑element discretizations, and convergence studies confirm that the error between the simulated field and the target pattern decays exponentially with the feedback gain, provided the gain remains within stability limits derived from the Lyapunov analysis.

The discussion emphasizes several implications. First, the framework unifies pattern generation, detection, and control within a single set of continuum equations, eliminating the need for separate sensing and actuation modules. Second, the universal approximation theorem suggests that any dynamical behavior—biological, chemical, or mechanical—can be embedded in a suitably engineered continuous medium, opening avenues for “programmed matter” where material properties are designed to execute arbitrary dynamical algorithms. Third, the methodology is not restricted to biochemical systems; it can be applied to multiphase fluids, soft solids, and even granular media, provided the constituent fields can be described by continuous concentration‑like variables.

Nevertheless, the authors acknowledge limitations. The feedback function σ and gain κ_i must be tuned, which may be nontrivial for real‑world materials where physical actuation is limited. Parameter identification for D_i, Φ_i, and R_i from experimental data remains an open inverse problem. Moreover, high‑dimensional PDE simulations are computationally intensive; future work will explore model‑order reduction, machine‑learning‑based surrogate models, and experimental validation in microfluidic and tissue‑culture platforms.

In conclusion, the paper delivers a mathematically rigorous, broadly applicable set of motion equations capable of recognizing and reproducing arbitrary spatiotemporal patterns. By proving that these equations can approximate any finite‑dimensional dynamical system, the authors provide a powerful theoretical tool for the design of adaptive, self‑organizing materials and for the systematic study of pattern formation across biology, chemistry, and materials science.