ProtoMech: Modular Mechanochemical Systems

• ProtoMech is a modular framework integrating mechanical, chemical, and computational primitives to achieve emergent, reprogrammable functionalities.
• It underpins diverse applications such as high-density mechanical memory, mechanochemical molecular swimmers, sparse protein circuit tracing, and coupled cell simulations.
• Design implications emphasize scalability, robustness, and precise energy landscape engineering for reliable actuation and dynamic performance.

ProtoMech refers to a class of systems, frameworks, or devices where fundamental processes are governed by coupled mechanical and chemical, computational, or structural mechanisms. Across recent scientific literature, "ProtoMech" designates foundational architectures for high-density reprogrammable mechanical memory, mechanochemical molecular swimmers, interpretable protein model circuit tracing, and spatial simulation of chemo-mechanical cell models. The unifying principle is the integration of discrete, modular primitives—mechanical, molecular, or computational—that realize emergent functions through carefully engineered interactions, state transitions, and information transfer.

1. Modular Architectures for Information Storage and Processing

ProtoMech metastructures, as detailed in (Li et al., 2024), are constructed from planar tessellations of "cube-and-hinge" unit cells. Each block comprises four 2×2 cube units (20 mm per side), linked by elastic rotational hinges with torsional stiffness $k$. The independent bistability of local struts arises from hinge rotations $\theta_x$, $\theta_y$, introducing multiple discrete mechanical states (bits) within each block. Bistable transitions occur at bifurcation points (e.g., $\theta_\mathrm{bif}=45^\circ$) in the configuration space, enabling reliable out-of-plane snap-through. For an $m\times n$ array, $5mn$ independently deformable struts yield $2^{5mn}-1$ discrete states (excluding the trivial all-down configuration), realized at an areal bit density given by

$$\rho_\mathrm{binary} = \frac{5}{8000}\;\mathrm{bits}/\mathrm{mm}^3 = 6.25\times 10^{-4}\;\mathrm{bits}/\mathrm{mm}^3.$$

Scaling to 1 mm cubes increases density to $5\;\mathrm{bits}/\mathrm{mm}^3$.

2. Mechanochemical Devices and Kinetic Modeling

In synthetic mechanochemical molecular swimmers ("ProtoMech" molecules) (Golestanian, 2010), information and work are coupled through stochastically cycling enzymatic modules. A canonical design utilizes a three-sphere swimmer: beads connected by flexible linkers, with catalyst-functionalized head and tail, and a permanently charged center. Complementary ionic reactions temporally load charge on the terminal beads, inducing sequential electrostatic deformations. The minimal chemical cycles are

$$F + h \xrightleftharpoons[k_{-f}]{k_f} hF \xrightarrow{k_4} hQ^- + G^+ \xrightarrow[k_{-5}]{k_5} h + Q^- + G^+,$$

$\theta_x$0

With six possible mechanochemical states in a cyclic network, steady-state dynamics are governed by a six-dimensional master equation, with net current $\theta_x$1. Geometric cycles in the $\theta_x$2 space (left/right linker extensions) yield propulsion at low Reynolds number, with average velocity following a Michaelis–Menten law: $\theta_x$3 where $\theta_x$4 is the area enclosed by the conformation cycle, $\theta_x$5 is rest length, $\theta_x$6 is bead diameter (Golestanian, 2010).

3. Sparse Circuit Tracing in Protein Models

ProtoMech in computational protein modeling (Tsui et al., 12 Feb 2026) introduces a "Cross-Layer Transcoder" (CLT) architecture, systematically extracting sparse, interpretable latent circuits from pretrained protein LLMs (pLMs). The CLT encodes the residual stream $\theta_x$7 at each transformer layer $\theta_x$8 into a strongly sparse latent $\theta_x$9 (using TopK hard sparsity, $\theta_y$0), and decodes jointly across layers to reconstruct true MLP outputs $\theta_y$1. The loss is a combination of mean-squared error and an auxiliary term to discourage dead units: $\theta_y$2 Compressed circuits are discovered by attribution-driven greedy selection: the minimal set $\theta_y$3 of latents retaining at least $\theta_y$4 (family classification, $\theta_y$5) or $\theta_y$6 (fitness, $\theta_y$7) of task accuracy using only $\theta_y$8 of total latents (Tsui et al., 12 Feb 2026).

Steering and interpretability experiments demonstrate alignment between circuit activity and biochemical motifs (e.g., kinase ATP-binding pockets, HRD motifs, Rossmann fold). ProtoMech-based circuit steering outperforms baselines for high-fitness protein design in 71% of high-throughput assays.

4. Simulation Frameworks for Coupled Chemo-Mechanical Systems

Spatial simulation platforms inspired by ProtoMech, such as Mechanica (Somogyi et al., 2017), provide a mesh-free, Lagrangian modeling environment for biological cells and tissues under coupled chemical and mechanical processes. Models consist of objects (stateful particles, links, regions, fields) and processes (continuous or discrete), specified by a dedicated modeling language. Continuous variables evolve via ODEs: $\theta_y$9 where $\theta_\mathrm{bif}=45^\circ$0 is the stoichiometry matrix, and $\theta_\mathrm{bif}=45^\circ$1 encodes user-defined rate laws. Particle mechanics are governed by soft DPD (dissipative particle dynamics) force balances. Dynamic creation/destruction, SBML-style guards for discrete processes, and efficient code generation (CVODE, C/AST compilation) support simulation of active cell migration, chemotaxis, and adhesion (Somogyi et al., 2017).

5. Energy Landscapes and Actuation in Mechanical ProtoMech

The ProtoMech mechanical bit architecture operates via local double-well energy landscapes. Under kinematic pre-stretch, the elastic potential for out-of-plane displacement $\theta_\mathrm{bif}=45^\circ$2 is

$\theta_\mathrm{bif}=45^\circ$3

with energy barrier $\theta_\mathrm{bif}=45^\circ$4. Magnetic reprogramming targets each bit independently, actuating snap-through when magnetic force $\theta_\mathrm{bif}=45^\circ$5 exceeds the critical threshold. Empirically, operation is robust to external pressures $\theta_\mathrm{bif}=45^\circ$62.8 kPa, lateral displacements $\theta_\mathrm{bif}=45^\circ$7 mm, and exhibits cycle lifetimes $\theta_\mathrm{bif}=45^\circ$8 without degradation (Li et al., 2024). Thermal retention is essentially infinite ($\theta_\mathrm{bif}=45^\circ$9).

6. Performance, Scalability, and Design Implications

ProtoMech devices and methodologies achieve a combination of high information density, modularity, and robustness:

• Information Storage: Reprogrammable metastructures yield densities of up to $m\times n$0 bits/mm$m\times n$1 at the microscale (5-level encoding) (Li et al., 2024).
• Dynamic Functionality: Mechanochemical molecular ProtoMech achieves velocities up to $m\times n$2 m/s for cycle rates $m\times n$3 s$m\times n$4 at $m\times n$5 nm, supporting design with scalable substrate and enzyme pairs (Golestanian, 2010).
• Circuit Interpretability: Protein model ProtoMech circuits achieve up to $m\times n$6 performance recovery with full CLT, and $m\times n$7 from $m\times n$8 of latents (Tsui et al., 12 Feb 2026).
• Simulation Capacity: Mesh-free Lagrangian ProtoMech frameworks efficiently simulate $m\times n$9–$5mn$0 particles and $5mn$1 chemical species in real-time scenarios (Somogyi et al., 2017).

Key design recommendations include matching actuation energies to barrier heights, optimizing geometric cycle areas for maximum output, and exploiting architectural independence of modular units for stability and scalability.

7. Functional Motifs and Biological Correspondence

ProtoMech circuits in protein modeling demonstrate correspondence between computationally discovered latent units and biochemically verified structures. Examples include arginine detectors and ATP-binding motifs in kinases, Rossmann folds in binding domains, and context-specific hydrophobic core detection for stability in engineered variants (Tsui et al., 12 Feb 2026). In mechanochemical swimmers, explicit conformational cycles are mapped to directional propulsion via non-reciprocal geometric motion in the conformational plane (Golestanian, 2010). Mechanical ProtoMech structures guarantee bit isolation and resilience via localized compatibility and high local stiffness (Li et al., 2024).

In summary, ProtoMech defines a modular framework for physically and computationally encoding, processing, and controlling information through coupled mechanical, chemical, or computational primitives, with demonstrable applications in memory, molecular propulsion, interpretability of protein models, and simulation of active matter.