Input-to-state stability-based chemical reaction networks composition for molecular computations

Chuanhou Gao; Denis Dochain; Di Fan; Renlei Jiang; Yuzhen Fan

REVIEW 3 major objections 4 minor 29 references

Input-to-state stability-based chemical reaction networks composition for molecular computations

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper establishes that two mass-action chemical reaction computers can be composed end-to-end and jointly compute the composite function, provided the downstream subsystem is input-to-state stable with respect to its equilibrium.

desk verdict A useful ISS-based cascade framework for CRN composition that overclaims generality: it only works as stated when downstream inputs are catalysts. read the letter →

arxiv 2506.12056 v1 pith:VXA3U5M4 submitted 2025-05-30 q-bio.MN

Renlei Jiang , Yuzhen Fan , Di Fan , Chuanhou Gao , Denis Dochain This is my paper · ORCID

classification q-bio.MN MSC 37N2592C4293D25

keywords chemicalreactionnetworksmass-actionkineticsmolecularcomputationcomposabilityinput-to-statestabilitydynamiclayer-by-layersigmoidfunction

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that molecular computers built from mass-action chemical reaction networks can simply be chained, with the output concentrations of one network feeding directly into the next, and that the chained system will compute the composite function provided the downstream network is input-to-state stable around its target equilibrium. If the claim is correct, layer-by-layer molecular computation no longer needs chemical oscillators to sequence the layers, and errors from finite clock phases disappear. The paper formalizes dynamic composability, proves sufficient conditions in terms of input-to-state stability, and demonstrates the approach on sigmoid and cube-root computations. This matters because complex biomolecular computations are currently built with many oscillator species and accumulate phase-timing errors; a stability-based composition rule would let designers assemble networks from verified elementary modules.

What carries the argument

The load-bearing object is the input-to-state stability estimate for the downstream output subsystem, Definition 2.7: $$\|y_2(t)-\bar y_2\| \le \$\beta$(\|y_2(0)-\bar y_2\|,t) + \gamma\!\left(\sup_{0\le \tau\le t}\|x_2(\tau)-\bar y_1\|\right).$$ The class-KL term $\beta$ forgets the downstream initial condition, and the class-K gain $\gamma$ converts the upstream output's deviation from $\bar y_1$ into an eventual output error. The proof of Theorem 3.3 splits time at $t/2$: the $\beta$ term has decayed by then, and since $y_1(t)\to \bar y_1$, the input-error term vanishes as $t\to\infty$. A second workhorse is the special output dynamics $\dot y = p(x) - q(x)y$ with $q(x)>0$, whose explicit solution converges to $p(\bar x)/q(\bar x)$; this single linear structure covers several examples and is the content of Proposition 3.2.

What would settle it

Integrate the Example 3.2 system $\dot x=0$, $\dot y=x+1-y$, $\dot z=y-z^3$ with a fixed $x_0$ and initial $z_0$ ranging from $1$ to $10^4$, and fit the smallest gain function that makes the claimed ISS bound hold. If the required gain grows with $z_0$, the uniformity assumption fails and Theorem 3.3's sufficient condition is not satisfied by that example as written.

Watch

Extended reading notes

Core claim

On its own terms, the paper claims that composability of molecular computers is a stability property, not a scheduling problem. Define a mass-action chemical reaction computer as a mass-action system with designated input and output species, computing a function when the output concentrations converge to the function of the initial inputs. Theorem 3.3 says that if the first computer's output $y_1(t)$ converges to $\bar y_1$, and the second computer's output dynamics $\dot y_2 = g_2(x_2, y_2)$ is input-to-state stable with respect to $(\bar y_1, \bar y_2)$, then the naively coupled system, with all reactions running together, satisfies $(y_1,y_2) \to (\bar y_1,\bar y_2)$. By Theorem 3.1, the composed network therefore computes $\sigma_2 \circ \sigma_1$. This is a sufficient condition, and the paper also gives easier-to-check sufficient conditions: output dynamics of the linear form $\dot y = p(x) - q(x)y$ with $q(x)>0$, the existence of an ISS-Lyapunov function, or global Lipschitz continuity together with global exponential stability. The worked examples include the sigmoid $1/(1+e^{-x_0})$ built from $e^{-x}$ and $1/(1+y)$, and $\sqrt[3]{x+1}$ built from $x+1$ and the cube root.

Load-bearing premise

The result depends on the downstream layer's error bound being uniform: the same shrinking-and-gain functions must work for every starting concentration and every bounded input, and the paper's worked example asserts, rather than proves, the boundedness constant that makes this true.

Editorial extensions

If this is right

The composition result is iterative: if every adjacent pair in a chain of $p$ mass-action chemical reaction computers is dynamically composable, the full chain dynamically computes $\sigma_p\circ\cdots\circ\sigma_1$ (Corollary 3.1 and Corollary 3.2).
The linear-output sufficient condition $\dot y = p(x) - q(x)y$ with $q(x)>0$ gives a direct design recipe: any network whose output species obey this equation can be placed after any network whose output converges, without adding clock species.
The finite-time error of the composed sigmoid computation decays as $C_2 e^{-(1-\varepsilon)T} + C_3 e^{-T}$, which the paper shows is faster than the oscillator-based method's $C_1 e^{-T/2}$ for the same readout time $T$; accuracy therefore improves without needing long clock phases.
Functions outside the reach of composable rate-independent chemical reaction networks, such as the sigmoid and $\sqrt[3]{x+1}$, become computable by mass-action composition, extending the class of functions molecular computers can reliably implement.
The examples show that coupled dynamics which at first appears to violate the desired layer-by-layer ordering can still be certified as dynamically composable, so direct cohabitation of reactions is acceptable when the ISS condition holds.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

An extension the paper leaves implicit: the same ISS criterion could serve as a compiler-side certificate, decompose a target function, verify each elementary layer's ISS-Lyapunov function, then assemble, giving the composable fundamental elementary library the paper names as future work.
A structural extension the paper does not treat: the cascade proof should carry over to multiple upstream outputs feeding one downstream layer when a vector-valued ISS gain exists, while feedback interconnections would need a small-gain argument, a neighboring problem rather than a result claimed here.
A testable prediction beyond the paper: in a DNA implementation of the chained sigmoid network, readout error at time $T$ should follow the paper's bound with the rate governed by the downstream ISS gain rather than by clock phasing; measuring error at several times $T$ would separate the two mechanisms.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick read on this one: the paper has a genuinely useful core idea—using ISS as a sufficient condition for cascading mass-action CRNs without oscillators—but the general composition claim is overstated because the framework only works when the downstream input species are pure catalysts, and the paper never says that.

What's actually new: extending composability from rate-independent CRNs (Chalk et al.) to mass-action systems, defining dynamic composability, and proving a cascade stability theorem (Thm 3.3) that essentially says: if the second module is ISS w.r.t. its equilibrium and the first module's output converges, then the composition converges. That is standard control theory, but the adaptation to CRNs and the worked examples (sigmoid, cube root) are nice. The error comparison in Thm 4.2 vs. oscillator-based sequential computation is original and well executed; it gives a quantitative statement that the ISS approach converges faster.

The soft spots. First, Definition 3.1 defines composition as the union of reaction sets, but the "dynamically composable" definition (Def 3.2) analyzes a cascade where the downstream input dynamics are frozen, ˙x2=0. Those two notions only coincide when downstream input species are pure catalysts. The paper never states this restriction. The counterexample is simple: C1 computes y=x via X→X+Y, Y→∅, and C2 computes z=y via Y→Z, Z→∅. The cascade predicts z→x, but the union dynamics gives ˙y=x−2y, so z→x/2. That undercuts the claimed generality of Theorem 3.1. The fix is either to restrict the framework to catalytic downstream inputs (where f2=0) or to extend the definition to include f2 and give conditions for when the cascade approximates the union. Second, there is a smaller definitional issue: the composed input set should be X1 ∪ (X2 \ Y1), not X1 ∪ X2 as written. Third, the ISS proof in Example 3.2 uses a uniform bound C that may depend on the initial condition; the claim is probably true by a standard ISS-Lyapunov argument, but as written the γ function is not shown to be initial-state independent.

Bottom line: the central idea is sound and the examples are convincing for the catalytic case, but as written the paper claims more generality than it proves. Worth sending to peer review with a request for major revision to fix the composition definition and state the catalyst assumption.

Referee Report

3 major / 4 minor

Summary. The paper proposes a framework for composing mass-action chemical reaction networks that act as analog computers. It defines mass-action chemical reaction computers (msCRCs), dynamic computation, and dynamic composability, and gives sufficient conditions, based on input-to-state stability (ISS), under which the cascade of two subsystems computes the composition of their individual functions. The central result, Theorem 3.3, states that if the output dynamics of the downstream network is ISS with respect to its equilibrium and the upstream output converges, then the coupled cascade converges to the composite limit. Several examples compute functions such as the sigmoid, cube root, and quadratic-formula map, and the paper compares its method with rate-independent CRN composition and oscillator-based sequential computation, including an explicit convergence-rate comparison in Theorem 4.2.

Significance. If the central claim were valid for the general class of mass-action CRNs, the paper would extend composability theory from rate-independent networks to rate-dependent systems and offer a principled alternative to oscillator-based phase separation. The ISS-based sufficient conditions are a sensible idea, and the explicit error-rate comparison with oscillator methods is a strength. However, the general claim is currently not established: the composition defined as a union of reaction sets differs from the cascade analyzed by the ISS theorems whenever the downstream input species participate non-catalytically, and this restriction is nowhere stated. All worked examples happen to use downstream inputs that are pure catalysts, but the framework as written is not restricted to that case. With the catalytic-input restriction made explicit and the definitions repaired, the paper would provide a correct and useful special-case framework.

major comments (3)

[Definitions 3.1 and 3.2, Theorem 3.3 and its proof] The composition defined in Definition 3.1 as the union of reaction sets R1 ∪ R2 is not the system analyzed in Definition 3.2 and Theorem 3.3. The latter analyze the cascade (19), in which the downstream input-species dynamics f2 is set to zero and no stoichiometric contribution of C2's reactions to the shared species Y1 = X2 appears. This is exact only when f2 ≡ 0, i.e., when every input species of C2 is a catalyst. The proof of Theorem 3.1 states without support that 'it takes f2 = 0'; this is the missing assumption, not a consequence of dynamic composability. For a downstream network containing Y → Z, the union composition gives y1-dynamics with an extra consumption term, while (19) omits it. Concretely, with C1: X → X + Y, Y → ∅ and C2: Y → Z, Z → ∅, the cascade model predicts z → x0, but the union ODE is ydot = x - 2y, zdot = y - z, whose limit is z → x0/2. The theorems and all examples must either explicitly restrict to cascades with catalytic downstream inputs or analyze the full union ODE.
[Definition 3.1] The input species set of the composite C2∘1 is defined as X1 ∪ X2. When Y1 = X2, this makes the intermediate species Y1 an input species of the composite even though its concentration is internally produced by C1. Under Definition 2.9, the computed function is evaluated at the initial input concentrations, so the composite as defined would be required to compute σ2∘σ1 from the initial concentrations of X1 and Y1 jointly, not from x1(0) alone. The proof of Theorem 3.1, however, uses only x1(0). The input set should be X1 ∪ (X2 \ Y1), or simply X1 when Y1 = X2, and the statement of Theorem 3.1 should be reconciled with the actual initial-state dependence.
[Example 3.2, Eq. (26) and following paragraph] The ISS certificate for z_dot = y - z^3 is incomplete. The proof bounds |z - zbar| by a constant C using the assertion that z is uniformly bounded when y is bounded, but no argument shows that C can be chosen independently of the initial condition z(0) and of the input bound sup|y - ybar|. In Definition 2.7, beta absorbs initial-state dependence but gamma must not depend on the initial state; if C depends on z(0), then gamma(s) = sqrt(2 C s / alpha) violates the definition. Since this example is the main demonstration of Theorem 3.3, the application of Theorem 3.3 to this example is not justified as written; a uniform bound independent of the initial state, or a different ISS proof producing a genuine class-K gain, is needed.

minor comments (4)

[Section III-A] The phrase 'design a msCSC' should read 'design a msCRC'.
[Appendix A, proof of Proposition 3.2] The coupled system in the proof reuses the symbol x for the upstream output y1, writing x(0) = y1_0 and x_dot = g1(x1, x); this notation is confusing and should be replaced by y1 throughout.
[Theorem 3.1] The word 'funtions' in the theorem statement is a typo for 'functions'.
[Section IV-B, Fig. 4] The caption of Fig. 4 does not identify which curve corresponds to the oscillator method and which to the proposed method; the curves should be labeled directly in the figure or in the caption.

Circularity Check

1 steps flagged · score 2.0 of 10

Minor definitional circularity in Theorem 3.1; the ISS-based sufficient conditions are derived independently and are not circular.

self definitional [Definition 3.2 and Theorem 3.1, Section III-B]
"Definition 3.2: 'If the solution (output components) of the coupled system ... satisfies lim_{t→∞} (y1(t), y2(t)) = (¯y1, ¯y2), then the two msCRCs are said to be dynamically composable.' Theorem 3.1: 'Then their composition C2◦1 is a dynamic computation of σ2 ◦ σ1 ... if C1 and C2 are dynamically composable.'"

The defining condition of 'dynamically composable' already asserts the convergence of the cascade to (¯y1, ¯y2). Since C1 is a dynamic computation of σ1, ¯y1=σ1(x1_0); since the isolated C2 in (18) has x2(0)=¯y1 and computes σ2, ¯y2=σ2(¯y1). Hence the conclusion of Theorem 3.1, namely lim y2=σ2(σ1(x1_0)), is exactly the condition in Definition 3.2. The proof only invokes Definition 3.2, so the theorem restates the definition rather than deriving new content. The substantive sufficient conditions (Proposition 3.2, Theorem 3.3) are independent.

full rationale

The paper's central contribution is the ISS-based sufficient conditions for composition (Proposition 3.2, Theorem 3.3, Corollary 3.2), and these are proved from the ISS estimate and standard cascade/L'Hopital arguments rather than from the conclusion being assumed. No parameters are fitted to data, no benchmark is tuned, and no non-external uniqueness theorem is imported from the authors' prior work. The only definitional circularity is Theorem 3.1, which is a restatement of Definition 3.2: dynamic composability is defined as the cascade converging to the individual limiting outputs, so the theorem that composable dynamic computations compose is true by definition. This is not load-bearing for the ISS conditions. A separate correctness gap, not a circularity, is that Definition 3.2 analyzes C2 only through ˙x2=0 and ˙y2=g2(x2,y2), so the coupled system (19) matches the union composition of Definition 3.1 only when C2's input species are catalysts; the paper does not state this restriction, and the skeptic's counterexample with a consuming downstream input is not addressed. Another non-circular correctness issue is that the uniform bound C in Example 3.2 is asserted rather than shown independent of the initial condition, so the claimed ISS gain may depend on the initial state. These gaps affect the stated generality but do not make the derivation circular. Overall circularity is therefore minimal.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

No new particles, forces, or conserved quantities are introduced. The framework relies on standard ODE theory, ISS theory, and the definition of dynamic computation. The examples use hand-chosen rate constants, but the general theorems do not depend on their specific values.

free parameters (1)

Example rate constants = 1, 1/2, 2, 4
Chosen by hand to simplify the example ODEs; the framework permits arbitrary positive constants, so these are not fitted to data.

assumptions (3)

standard math L'Hopital's rule is used in the proofs of Proposition 3.2 and Theorem 4.2 to evaluate limits of ratios.
Invoked in the proof of Proposition 3.2 (Section III-B) and Theorem 4.2 (Appendix -A).
domain assumption Mass-action kinetics produces locally Lipschitz polynomial vector fields, ensuring existence and uniqueness of solutions and well-defined limiting behavior.
Assumed throughout Definitions 2.4-2.9 and used implicitly in all examples.
standard math Input-to-state stability theory as in Khalil, including the ISS-Lyapunov sufficient condition (Lemma 0.1) and the global Lipschitz condition (Lemma 0.2), is imported without proof.
Lemmas 0.1 and 0.2 are cited from [29] in the proof of Proposition 3.4.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Input-to-state stability-based chemical reaction networks composition for molecular computations." pith.science (2026). https://pith.science/paper/VXA3U5M4

@misc{pith2026250612056,
  author       = {Pith},
  title        = {Pith review of: Input-to-state stability-based chemical reaction networks composition for molecular computations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VXA3U5M4}},
  note         = {Machine review of arXiv:2506.12056}
}

read the original abstract

Molecular computation based on chemical reaction networks (CRNs) has emerged as a promising paradigm for designing programmable biochemical systems. However, the implementation of complex computations still requires excessively large and intricate network structures, largely due to the limited understanding of composability, that is, how multiple subsystems can be coupled while preserving computational functionality. Existing composability frameworks primarily focus on rate-independent CRNs, whose computational capabilities are severely restricted. This article aims to establish a systematic framework for composable CRNs governed by mass-action kinetics, a common type of rate-dependent CRNs. Drawing upon the concepts of composable rate-independent CRNs, we introduce the notions of mass-action chemical reaction computers (msCRCs), dynamic computation and dynamic composability to establish a rigorous mathematical framework for composing two or more msCRCs to achieve layer-by-layer computation of composite functions. Further, we derive several sufficient conditions based on the notions of input-to-state stability (ISS) to characterize msCRCs that can be composed to implement desired molecular computations, thereby providing theoretical support for this framework. Some examples are presented to illustrate the efficiency of our method. Finally, comparative results demonstrate that the proposed method exhibits notable advantages in both computational ability and accuracy over the state-of-the-art methods.

Figures

Figures reproduced from arXiv: 2506.12056 by the authors.

**Figure 1.** A schematic diagram on chemical oscillator with four species, in which O1 and O3 are used to control two reaction networks. proposed to use oscillating signals as catalysts, which are generated by Oi + Oi+1 k0 −→ 2Oi+1 (1 ≤ i ≤ r − 1), Or + O1 k0 −→ 2O1, (14) with dynamics    o˙1 = k0(o1or − o1o2), · · · o˙r = k0(or−1or − oro1), (15) where Oi denotes the clock signal species. For two networks, there needs to pr… view at source ↗

**Figure 2.** The evolution of the concentration of each species in (29) with three kinds of different initial values: (a) x1(0) = 1.0 > x2(0) = 0.6, y(0) = 0, z(0) = 1.0; (b) x1(0) = 0.7 < x2(0) = 0.9, y(0) = 0, z(0) = 0; (c) x1(0) = x2(0) = 0.8, y(0) = 0, z(0) = 0.5 (Under this condition, the curves of x1(t) and x2(t) overlap). As an example of the sigmoid function z = 1 1+e−x , the network in (16) is able to perform the comput… view at source ↗

**Figure 4.** Errors comparison of computing sigmoid function between the oscillator method and our method with x0 = 1.8, y0 = 1, z0 = 0. V. CONCLUSIONS In this paper, we aim at following the composition rule of functions to design CRNs to implement molecular computations. By borrowing the concepts of msCRC and rateindependent CRNs composition [18], we define the notions of msCRC, dynamic computation and dynamic composability.… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 29 canonical work pages

[1]

A synthetic protein-level neural network in mammalian cells,

Z. Chen, J. M. Linton, S. Xia, X. Fan, D. Yu, J. Wang, R. Zhu, and M. B. Elowitz, “A synthetic protein-level neural network in mammalian cells,” Science, vol. 386, no. 6727, pp. 1243–1250, 2024

work page 2024
[2]

Programmable probiotics for detection of cancer in urine,

T. Danino, A. Prindle, G. A. Kwong, M. Skalak, H. Li, K. Allen, J. Hasty, and S. N. Bhatia, “Programmable probiotics for detection of cancer in urine,” Science Translational Medicine , vol. 7, no. 289, pp. 289ra84–289ra84, 2015. 12 IEEE TRANSACTIONS AND JOURNALS TEMPLATE

work page 2015
[3]

Microbial engineering for the production of advanced biofuels,

P. P. Peralta-Yahya, F. Zhang, S. B. Del Cardayre, and J. D. Keasling, “Microbial engineering for the production of advanced biofuels,”Nature, vol. 488, no. 7411, pp. 320–328, 2012

work page 2012
[4]

Chemical reaction network structure and the stability of complex isothermal reactors—i. the deficiency zero and deficiency one theorems,

M. Feinberg, “Chemical reaction network structure and the stability of complex isothermal reactors—i. the deficiency zero and deficiency one theorems,” Chemical Engineering Science , vol. 42, no. 10, pp. 2229– 2268, 1987

work page 1987
[5]

Persistence and permanence of mass-action and power-law dynamical systems,

G. Craciun, F. Nazarov, and C. Pantea, “Persistence and permanence of mass-action and power-law dynamical systems,” SIAM Journal on Applied Mathematics, vol. 73, no. 1, pp. 305–329, 2013

work page 2013
[6]

Bistability and oscillations in chemical reaction networks,

M. Domijan and M. Kirkilionis, “Bistability and oscillations in chemical reaction networks,” Journal of Mathematical Biology , vol. 59, no. 4, pp. 467–501, 2009

work page 2009
[7]

Saddle-node bifurcations in biochemical reaction networks with mass action kinetics and application to a double-phosphorylation mechanism,

C. Conradi, D. Flockerzi, and J. Raisch, “Saddle-node bifurcations in biochemical reaction networks with mass action kinetics and application to a double-phosphorylation mechanism,” in 2007 American Control Conference, pp. 6103–6109, IEEE, 2007

work page 2007
[8]

Molecular computation of solutions to combinatorial problems,

L. M. Adleman, “Molecular computation of solutions to combinatorial problems,” Science, vol. 266, no. 5187, pp. 1021–1024, 1994

work page 1994

Show all 29 references

[9]

Computational functions in biochemical reaction networks,

A. Arkin and J. Ross, “Computational functions in biochemical reaction networks,” Biophysical Journal, vol. 67, no. 2, pp. 560–578, 1994

1994
[10]

Computing algebraic functions with biochemical reaction networks,

H. Buisman, H. M. ten Eikelder, P. A. Hilbers, and A. M. Liekens, “Computing algebraic functions with biochemical reaction networks,” Artificial Life, vol. 15, no. 1, pp. 5–19, 2009

2009
[11]

Strong turing com- pleteness of continuous chemical reaction networks and compilation of mixed analog-digital programs,

F. Fages, G. Le Guludec, O. Bournez, and A. Pouly, “Strong turing com- pleteness of continuous chemical reaction networks and compilation of mixed analog-digital programs,” in Computational Methods in Systems Biology: 15th International Conference, CMSB 2017, Darmstadt, Ger- m...

2017
[12]

Dna as a universal substrate for chemical kinetics,

D. Soloveichik, G. Seelig, and E. Winfree, “Dna as a universal substrate for chemical kinetics,” Proceedings of the National Academy of Sciences, vol. 107, no. 12, pp. 5393–5398, 2010

2010
[13]

Scaling up molecular pattern recognition with dna-based winner-take-all neural networks,

K. M. Cherry and L. Qian, “Scaling up molecular pattern recognition with dna-based winner-take-all neural networks,” Nature, vol. 559, no. 7714, pp. 370–376, 2018

2018
[14]

Nonlinear decision-making with enzymatic neural networks,

S. Okumura, G. Gines, N. Lobato-Dauzier, A. Baccouche, R. Deteix, T. Fujii, Y . Rondelez, and A. J. Genot, “Nonlinear decision-making with enzymatic neural networks,” Nature, vol. 610, no. 7932, pp. 496–501, 2022

2022
[15]

Crn++: Molecular pro- gramming language,

M. Vasi ´c, D. Soloveichik, and S. Khurshid, “Crn++: Molecular pro- gramming language,” Natural Computing, vol. 19, no. 2, pp. 391–407, 2020

2020
[16]

Chemical reaction networks for computing logarithm,

C. T. Chou, “Chemical reaction networks for computing logarithm,” Synthetic Biology, vol. 2, no. 1, p. ysx002, 2017

2017
[17]

Rate-independent computa- tion in continuous chemical reaction networks,

H.-L. Chen, D. Doty, and D. Soloveichik, “Rate-independent computa- tion in continuous chemical reaction networks,” inProceedings of the 5th Conference on Innovations in Theoretical Computer Science , pp. 313– 326, 2014

2014
[18]

Compos- able rate-independent computation in continuous chemical reaction networks,

C. Chalk, N. Kornerup, W. Reeves, and D. Soloveichik, “Compos- able rate-independent computation in continuous chemical reaction networks,” IEEE/ACM Transactions on Computational Biology and Bioinformatics, vol. 18, no. 1, pp. 250–260, 2019

2019
[19]

Sequestration and delays enable the synthesis of a molecular derivative operator,

C. C. Samaniego, J. Kim, and E. Franco, “Sequestration and delays enable the synthesis of a molecular derivative operator,” in 2020 59th IEEE Conference on Decision and Control (CDC) , pp. 5106–5112, IEEE, 2020

2020
[20]

Chemical mass-action systems as analog computers: Implementing arithmetic computations at specified speed,

D. F. Anderson and B. Joshi, “Chemical mass-action systems as analog computers: Implementing arithmetic computations at specified speed,” Theoretical Computer Science , vol. 1025, p. 114983, 2025

2025
[21]

Programming and training rate-independent chemical reaction net- works,

M. Vasi ´c, C. Chalk, A. Luchsinger, S. Khurshid, and D. Soloveichik, “Programming and training rate-independent chemical reaction net- works,” Proceedings of the National Academy of Sciences , vol. 119, no. 24, p. e2111552119, 2022

2022
[22]

On reaction network implementations of neural networks,

D. F. Anderson, B. Joshi, and A. Deshpande, “On reaction network implementations of neural networks,” Journal of the Royal Society Interface, vol. 18, no. 177, p. 20210031, 2021

2021
[23]

Automatic implementation of neural networks through reaction networks—part i: Circuit design and convergence analysis,

Y . Fan, X. Zhang, C. Gao, and D. Dochain, “Automatic implementation of neural networks through reaction networks—part i: Circuit design and convergence analysis,” IEEE Transactions on Automatic Control , 2025

2025
[24]

The computationally complete ant colony: Global coordination in a system with no hierarchy,

M. Lachmann and G. Sella, “The computationally complete ant colony: Global coordination in a system with no hierarchy,” in European Conference on Artificial Life , pp. 784–800, Springer, 1995

1995
[25]

Controlling the occurrence sequence of reaction modules through biochemical relaxation oscillators,

X. Shi, C. Gao, and D. Dochain, “Controlling the occurrence sequence of reaction modules through biochemical relaxation oscillators,” Auto- matica, vol. 176, p. 112261, 2025

2025
[26]

Automatic implementation of neural networks through reaction networks–part ii: Error analysis,

Y . Fan, X. Zhang, C. Gao, and D. Dochain, “Automatic implementation of neural networks through reaction networks–part ii: Error analysis,” arXiv preprint arXiv:2401.07077 , 2024

2024 arXiv
[27]

Smooth stabilization implies coprime factorization,

E. D. Sontag, “Smooth stabilization implies coprime factorization,” IEEE Transactions on Automatic Control , vol. 34, no. 4, pp. 435–443, 1989

1989
[28]

Rouche, P

N. Rouche, P. Habets, and M. Laloy, Stability theory by Liapunov’s direct method, vol. 4. Springer, 1977

1977
[29]

H. K. Khalil and J. W. Grizzle, Nonlinear Systems, vol. 3. Prentice hall Upper Saddle River, NJ, 2002

2002

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

A synthetic protein-level neural network in mammalian cells,

Z. Chen, J. M. Linton, S. Xia, X. Fan, D. Yu, J. Wang, R. Zhu, and M. B. Elowitz, “A synthetic protein-level neural network in mammalian cells,” Science, vol. 386, no. 6727, pp. 1243–1250, 2024

work page 2024

[2] [2]

Programmable probiotics for detection of cancer in urine,

T. Danino, A. Prindle, G. A. Kwong, M. Skalak, H. Li, K. Allen, J. Hasty, and S. N. Bhatia, “Programmable probiotics for detection of cancer in urine,” Science Translational Medicine , vol. 7, no. 289, pp. 289ra84–289ra84, 2015. 12 IEEE TRANSACTIONS AND JOURNALS TEMPLATE

work page 2015

[3] [3]

Microbial engineering for the production of advanced biofuels,

P. P. Peralta-Yahya, F. Zhang, S. B. Del Cardayre, and J. D. Keasling, “Microbial engineering for the production of advanced biofuels,”Nature, vol. 488, no. 7411, pp. 320–328, 2012

work page 2012

[4] [4]

Chemical reaction network structure and the stability of complex isothermal reactors—i. the deficiency zero and deficiency one theorems,

M. Feinberg, “Chemical reaction network structure and the stability of complex isothermal reactors—i. the deficiency zero and deficiency one theorems,” Chemical Engineering Science , vol. 42, no. 10, pp. 2229– 2268, 1987

work page 1987

[5] [5]

Persistence and permanence of mass-action and power-law dynamical systems,

G. Craciun, F. Nazarov, and C. Pantea, “Persistence and permanence of mass-action and power-law dynamical systems,” SIAM Journal on Applied Mathematics, vol. 73, no. 1, pp. 305–329, 2013

work page 2013

[6] [6]

Bistability and oscillations in chemical reaction networks,

M. Domijan and M. Kirkilionis, “Bistability and oscillations in chemical reaction networks,” Journal of Mathematical Biology , vol. 59, no. 4, pp. 467–501, 2009

work page 2009

[7] [7]

Saddle-node bifurcations in biochemical reaction networks with mass action kinetics and application to a double-phosphorylation mechanism,

C. Conradi, D. Flockerzi, and J. Raisch, “Saddle-node bifurcations in biochemical reaction networks with mass action kinetics and application to a double-phosphorylation mechanism,” in 2007 American Control Conference, pp. 6103–6109, IEEE, 2007

work page 2007

[8] [8]

Molecular computation of solutions to combinatorial problems,

L. M. Adleman, “Molecular computation of solutions to combinatorial problems,” Science, vol. 266, no. 5187, pp. 1021–1024, 1994

work page 1994