{"id":"b1351420-398d-44fc-8c66-19b16b6fb58e","arxiv_id":"2507.12699","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"An iterative leveling algorithm constructs detailed-balance equilibrium concentrations for athermic monomer-polymer systems, exponentially suppressing off-target species, and applies this to bound leak in DNA circuits.","lead":"This paper presents an iterative algorithm that assigns equilibrium concentrations to molecular complexes so that desired complexes are abundant and undesired ones are exponentially rare, in an enthalpy-free molecular setting. It also shows how this framework bounds 'leak' in DNA logic circuits and suggests a design change to make leak decay exponentially with system size.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 5.4's claim that x0∈(0,1) for any 0<c<1 is false: with S={A={a}} and B={a,a}, x0(A)=c+2c^2 exceeds 1 for c=0.9, so the theorem needs a small-c quantifier.","rationale":"The reader's weakest assumption was stability of S and the informal worst-case reaction claims in Section 8. Those are real concerns about the applicability of the framework, and the paper explicitly flags them. However, the most load-bearing flaw I find is sharper: Theorem 5.4 is false as stated because the mass-conservation condition x0=A·x is not guaranteed to lie in (0,1) for every c<1. The counterexample above is minimal and verifiable, and it directly contradicts a stated part of the central theorem. The fix is straightforward — quantify over sufficiently small c, which matches the intended dilute regime and still supports the leak bounds in Section 8, where c is a small mole fraction. Because the core algorithmic construction may still be correct after this revision, I do not recommend outright rejection; the paper should be conditionally accepted with the theorem corrected and the proof gaps (Case 2 of Theorem 5.4, Lemma 5.3's sketch, and the x0 bound) repaired. The reader already noticed the x0 issue in passing, so my concern partially overlaps with theirs, but they did not elevate it to the weakest assumption.","tokens_in":17230,"tokens_out":21283,"duration_ms":256032,"concrete_test":"Instantiate the counterexample: fix Ψ0={a}, S={A={a}} with μ(A)=1, and Ψ={A, B={a,a}}. Run Algorithm 1 (or compute by hand) to confirm μ̄(B)=2. For c=0.9, compute x0(a)=x_A+2x_B=0.9+2(0.81)=2.52; if x0∉(0,1), Theorem 5.4's requirement fails. Then repeat with c=0.4, where x0(a)=0.4+2(0.16)=0.72∈(0,1), confirming that the small-c version holds. This single computation settles whether the universal quantifier over c is valid or must be restricted.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Theorem 5.4 asserts that for any 0<c<1 there are monomer concentrations x0∈(0,1)^Ψ0 such that x_P=c^{μ̄(P)} is the equilibrium. This is false as stated because x0=A·x is forced by mass conservation, and Algorithm 1 imposes no bound keeping its entries below 1. Concrete counterexample: take Ψ0={a}, S={A={a}} with uniform μ(A)=1, and Ψ={A, B={a,a}}. S is stable: every canonical reaction has k/l=2 (e.g., 2A→B has k=2, l=1; nA→mB+rA has k=2m, l=m). Algorithm 1 assigns μ̄(B)=2. For c=0.9, x=(0.9,0.81) lies in (0,1)^Ψ, detailed balance holds, but x0(a)=x_A+2x_B=0.9+2(0.81)=2.52>1, contradicting x0∈(0,1)^Ψ0. The paper itself notes that mole fractions require 'the regime of less polymer than solvent' (Remark 3.2), so the correct theorem is 'for all sufficiently small c' or with an explicit bound; the universal quantifier over c is internally inconsistent with the model. This is a direct flaw in the central existence claim, independent of the informal worst-case reaction claims in Section 8. A related proof gap remains: Case 2 of Theorem 5.4 sums levelizing reactions β_P for every product P∈M2, but on-target products P∈S have no levelizing reaction, so the reduction to Case 1 is not defined for canonical reactions with on-target products.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies equilibrium concentrations in an athermic polymer-monomer model, where polymers are multisets of monomers and the free energy is purely entropic. Given a set S of on-target polymers with prescribed concentration exponents mu, the authors define canonical reactions, their imbalance k(alpha) and novelty l(alpha), and a stability condition k(alpha)/l(alpha) > 1. Algorithm 1 iteratively assigns extended exponents bar-mu to all polymers by levelizing canonical reactions. Theorem 5.4 claims that, for any 0 < c < 1, there exist monomer concentrations x0 in (0,1) such that setting each polymer P to c^{bar-mu(P)} minimizes the entropic free energy subject to mass conservation, with all off-target exponents strictly above 1. The paper then derives a bound based on TBN-stability (Corollary 7.2) and applies it to an AND gate and to translator cascades, arguing that leak concentrations scale polynomially in c and, in a modified parameter regime, exponentially in the redundancy parameter.","tokens_in":17576,"tokens_out":5162,"duration_ms":65783,"significance":"If the main theorem is correct, the levelizing construction gives a genuinely useful bridge between combinatorial TBN entropy-loss arguments and real-valued equilibrium concentrations, and the paper's explicit algorithm and Hilbert-basis treatment (Appendix A) are valuable contributions. The monotonicity lemma (Lemma 6.3) and the bound framework (Section 6) are also interesting and potentially reusable. However, the central existence theorem is currently false as stated because of the unrestricted quantifier over c, and the proof contains a genuine gap in Case 2. The application results in Section 8 are explicitly conditional on unproven worst-case reaction identifications. These issues materially affect the paper's main claims, so the significance is real but conditional on a substantial revision.","major_comments":[{"comment":"The statement 'for any 0 < c < 1' is false because mass conservation forces x0 = A·x, and the entries of x0 must lie in (0,1). A concrete counterexample is S = {A} with A = {a}, an off-target polymer B = {a,a}, and uniform mu(A) = 1. S is stable and Algorithm 1 assigns bar-mu(B) = 2. For c = 0.9, the configuration x_A = 0.9, x_B = 0.81 lies in (0,1)^Psi and satisfies detailed balance, but x0(a) = x_A + 2 x_B = 2.52, contradicting x0 in (0,1)^Psi0. The theorem needs a small-c quantifier or an explicit upper bound on c; this is not a presentation issue but a load-bearing error in the main existence result.","section":"Theorem 5.4 and Definition 2.1"},{"comment":"The reduction in Case 2 is undefined for canonical reactions with on-target products. The proof says to choose, for each polymer P in M2, a levelizing reaction beta_P that includes P as a product, and then sums M2[P] copies of beta_P. But if P belongs to the on-target set S, there is no levelizing reaction for P, because levelizing reactions are canonical reactions with l(alpha) != 0 that attain the minimum ratio, whereas on-target polymers are level 0 and are not assigned by any levelizing reaction. Thus the sum over P in M2 cannot be formed when M2 intersects S. The proof needs a separate argument for on-target products before the contradiction can be concluded.","section":"Theorem 5.4, proof, Case 2"},{"comment":"The proof of Lemma 5.3 is only a sketch and does not rigorously establish that every reaction is a combination of canonical reactions. The displayed derivation replaces alpha by a reaction involving M1 + M1' and M2 + M2', but it is not shown that repeating this procedure terminates in a reaction whose reactants all lie in S, nor that the final combined reaction is equivalent to the original one in the sense needed for detailed balance. Since Lemma 5.3 is the bridge between balancing canonical reactions and minimizing g(x) subject to mass conservation, this gap is load-bearing and requires a complete proof.","section":"Lemma 5.3"},{"comment":"The application bounds are not established because the worst-case canonical reactions are asserted without proof. The paper explicitly states in Section 8 that the authors 'claim to have identified the worst-case canonical reactions' and in the Discussion that 'the argument is informal.' Consequently the concrete leak bounds c^{1.5}, c^{1.33}, c^{4/3}, and c^{1/4+N/4} are conditional on unverified stability and worst-case assumptions. Either a proof of the worst-case identification must be supplied, or the statements must be reworded as conjectures or heuristic bounds.","section":"Section 8 and Corollary 7.2"}],"minor_comments":[{"comment":"The footnote in Remark 3.2 already acknowledges that mole-fraction concentrations require 'the regime of less polymer than solvent,' and this should be reconciled with the unrestricted 'any 0 < c < 1' quantifier in Theorem 5.4 and Corollary 7.2.","section":"Remark 3.2"},{"comment":"The sentence 'Append all polymers P in M2 that are not in hat(M2) to S_i and assigns bar-mu(P) = mu_i' has a subject-verb agreement issue and should be rephrased, for example as 'assign bar-mu(P) = mu_i to each such P.'","section":"Algorithm 1, line 12"},{"comment":"The notation hat(M2) is reused in Definition 5.1 for the level-dependent intersection with the union of previously levelized sets, while the same symbol is used in Definition 3.3 for M2 intersect S. The dependency on i should be made explicit, such as hat(M2)_i, to avoid confusion.","section":"Definitions 5.1 and 5.2"}],"recommendation":"major_revision","confidential_remarks":"The paper's main theorem appears salvageable by adding a small-c condition and repairing the Case 2 argument, but as submitted the universal quantifier over c is demonstrably false and the proof of Lemma 5.3 is incomplete. The authors' own Discussion admits that the worst-case reaction claims in Section 8 are informal, so the application sections should be clearly labeled as heuristic unless the missing verification is supplied."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe leveling algorithm is the real thing. The idea of assigning concentration exponents level by level, using the ratio of imbalance to novelty, is new and produces explicit upper bounds on off-target concentrations that the log-linear approach cannot easily provide. Corollary 7.2, connecting TBN entropy loss to concentration exponents, is clean and useful. And the translator cascade analysis—showing that increasing redundancy N alone does not arbitrarily suppress equilibrium leak—is a genuinely valuable corrective to the 'leakless' label. The Hilbert basis appendix is a thoughtful way around the infinite set of canonical reactions.\n\nThat said, the main theorem as stated is false. Theorem 5.4 says that for any 0<c<1 there are monomer concentrations x0 in (0,1) such that the configuration with each polymer at c^{μ̄(P)} is the equilibrium. But x0 is forced by mass conservation: x0 = A·x. Take S={A}, μ(A)=1, B=2A, with Ψ={A,B}. Algorithm 1 assigns μ̄(B)=2, and then x0(A)=c+2c^2, which exceeds 1 for any c>1/2. So the universal quantifier over c must be replaced by a small-c condition or an explicit bound on c. The paper actually cares about the dilute regime, so this is fixable, but it is a real error in the central existence claim.\n\nThe proof of Theorem 5.4 has another gap in Case 2. The reduction sums levelizing reactions β_P for every P in M2, but on-target products in M2 have no levelizing reaction, so the sum is not defined for canonical reactions whose products include on-target polymers. I suspect the case can be patched, but the proof as written is incomplete.\n\nSection 8's worst-case reaction claims are explicitly informal. The leak bounds for the AND gate and the translator cascade depend on the claim that the identified reactions minimize e/l or k/l, and the authors admit they have no proof. For the translator, the qualitative point that novelty grows with N is likely robust, but the exact exponential bound should be treated as conditional on those unproven claims.\n\nThis paper deserves a serious referee. The framework is novel, the connection between TBN combinatorics and real-valued concentrations is valuable, and the flaws are localized to the theorem statement, one proof case, and the informal application claims. I would send it to peer review, requiring the c-quantifier fix and the Case 2 patch; the application section should either contain proofs or be explicitly labeled as conjectural. If you work on DNA leak, cite the leveling framework and the redundancy result, but verify the worst-case claims before relying on the bounds.","headline":"The leveling algorithm is a genuine contribution, but Theorem 5.4 overreaches with a false universal-c claim and the application section rests on unproven worst-case reactions.","tokens_in":18132,"tokens_out":4891,"would_cite":true,"duration_ms":53456,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92E20"],"pacs":[],"model":"deepseek-v4-flash","headline":"For any stable on-target polymer set, an iterative leveling algorithm assigns concentration exponents so that the configuration $c^{\\bar{\\mu}(P)}$ is the exact equilibrium, with every off-target polymer exponentially suppressed.","keywords":["equilibrium concentrations","athermic systems","detailed balance","thermodynamic binding networks","monomer-polymer model","leak suppression","DNA strand displacement","iterative leveling algorithm"],"falsifier":"Compute the finite Hilbert basis of canonical reactions for a proposed stable set $S$ and test $k(\\alpha)/l(\\alpha)>1$ on every basis element with $l(\\alpha)>0$; any ratio $\\le 1$ refutes the conclusion that all off-target exponents exceed 1. For the translator cascade, solve the detailed-balance equations at the specified fuel and waste concentrations and check whether the leak output is below $c^{1/4+N/4}$ and whether more than half the signal reaches the output layer; failure of either inequality would falsify the worst-case-reaction claim.","tokens_in":16993,"feed_emoji":"🧬","tokens_out":9417,"duration_ms":106723,"temperature":0.7,"pith_summary":"The paper asks whether equilibrium concentrations of assembled molecular complexes can be set by design rather than found by numerical simulation, and answers it in the athermic case where all interactions are enthalpy-neutral. It proves that from a chosen high-concentration set $S$, an iterative leveling algorithm extends the assignment to every polymer: at each step the newly introduced polymers receive the concentration exponent that balances the canonical reaction with the smallest imbalance-to-novelty ratio $k(\\alpha)/l(\\alpha)$. When $S$ is stable, meaning every canonical reaction that creates something new satisfies $k(\\alpha)/l(\\alpha)>1$, the resulting configuration is the minimum of the entropic free energy under mass conservation, and all off-target polymers have concentration exponents above 1, hence concentrations below the on-target baseline $c$. This turns combinatorial entropy-loss arguments in the Thermodynamic Binding Networks (TBN) model into explicit real-valued concentration bounds, and the paper demonstrates the bounds on a DNA AND gate and on a parameterized translator cascade, where it predicts exponentially decreasing leak once waste concentrations are tuned.","feed_headline":"One ratio sets every equilibrium concentration in athermic chemistry","feed_subtitle":"Iterative leveling turns entropy-loss arguments into exponential leak bounds for DNA systems.","key_machinery":"The load-bearing mechanism is the level-by-level construction over canonical reactions. A canonical reaction has all reactants in the on-target set; its $i$th-level imbalance $k_i(\\alpha)=\\bar{\\mu}(M_1)-\\bar{\\mu}(\\hat{M}_2)$ measures the exponent surplus on the known side, and its novelty $l_i(\\alpha)$ counts how many new off-target polymers the reaction introduces. The ratio $k_i(\\alpha)/l_i(\\alpha)$ is the effective price of creating new species: high imbalance makes products expensive, while high novelty makes them entropically cheap. Algorithm 1 sets $\\mu_i$ to the minimum such ratio, assigns that same exponent to every polymer first appearing in a minimizing reaction, and repeats until all polymers are levelized. Stability of $S$ is exactly the condition that the first such ratio is strictly greater than 1, forcing the exponent floor $\\mu_1>1$. The Hilbert-basis argument reduces the infinite set of canonical reactions to a finite generating set, making the minimum well-defined and the algorithm terminating, and Lemma 5.3 converts balance on canonical reactions into the true minimum of the entropic free energy.","core_discovery":"The central claim is Theorem 5.4: if $S$ is stable with concentration exponents $\\mu$, then Algorithm 1 produces extended exponents $\\bar{\\mu}$ such that for any $0<c<1$ there exist monomer concentrations $x_0$ with the configuration $x_P = c^{\\bar{\\mu}(P)}$ minimizing $g(x)=\\sum_{P\\in\\Psi} x_P(\\log x_P-1)$ subject to $A\\cdot x=x_0$, and every off-target polymer satisfies $\\bar{\\mu}(P)\\ge \\mu_1>1$. The proof shows every canonical reaction is balanced in that configuration: looking at the highest level among a reaction's products forces the level-defining minimum to contradict itself, and the opposite direction reduces to this case by summing the levelizing reactions of the products. A lemma shows that balancing canonical reactions is sufficient for a global free-energy minimum, because arbitrary reactions decompose into canonical ones. As a corollary, a TBN-stability-closed set with uniform on-target exponents has off-target concentrations bounded by $c^{\\min_\\alpha(e(\\alpha)/l(\\alpha))+1}$, where $e(\\alpha)$ is entropy loss and $l(\\alpha)$ is novelty of a canonical reaction.","pith_inferences":["The ratio-minimization view suggests a computational test for stability: enumerate the finite Hilbert basis of canonical reactions and check $k(\\alpha)/l(\\alpha)>1$ on that generating set, turning the informal worst-case arguments used for the applications into a verification algorithm.","The same leveling recursion can be read as a min-ratio path problem on the reaction cone, so efficient combinatorial optimization algorithms may yield the worst-case ratios and hence the concentration exponents without enumerating all reactions.","If the framework were extended to per-polymer free energies $\\Delta G_P$, the imbalance would likely become $\\bar{\\mu}(M_1)-\\bar{\\mu}(\\hat{M}_2)$ shifted by free-energy terms, giving a natural route toward toehold-mediated, non-athermic DNA systems.","The translator-cascade analysis implies a design principle: to make leak decay with a redundancy parameter, keep the novelty $l(\\alpha)$ of the worst leak reaction constant while growing its imbalance, which is achieved precisely by nonuniform on-target concentration exponents."],"forward_implications":["For any stable on-target set, equilibrium concentrations are obtained by the leveling recursion instead of by numerically minimizing the free energy, and each early stop of the algorithm yields a valid upper bound on every not-yet-levelized off-target polymer.","In TBN systems, any TBN-stability-closed set with uniform $\\mu=1$ gives explicit leak bounds of the form $c^{\\min(e/l)+1}$, so combinatorial entropy-loss statements become quantitative concentration guarantees.","For the DNA AND gate, the analysis gives leak $\\le c^{1.5}$ without inputs and $\\le c^{1.33}$ with one input present, so leak is polynomially suppressed as concentrations shrink.","For two translators in cascade, uniform fuel concentrations only tighten the leak bound toward $c^2$ as redundancy $N$ grows; setting waste at concentration $c$ and fuel at $2c$ instead makes the bound $c^{1/4+N/4}$, giving exponential leak suppression in $N$ while preserving a constant fraction of signal reaching the output.","All of these consequences hold for equilibrium concentrations under athermic, enthalpy-neutral conditions, the regime of saturated strongly bonded DNA domains."],"supporting_citations":[{"why":"supplies the free-energy function $g(x)$ whose minimizer, under mass conservation, defines equilibrium concentrations.","marker":"[5]"},{"why":"establishes that equilibrium minimizers satisfy detailed balance in mass-action systems, the condition the leveling algorithm enforces.","marker":"[10]"},{"why":"provides the chemical reaction network theory background on detailed-balance equilibria used to justify the free-energy formulation.","marker":"[8]"},{"why":"introduces Thermodynamic Binding Networks and the entropy/stability notions the paper connects to concentrations.","marker":"[6]"},{"why":"defines the leakless DNA strand displacement translator family and the entropy-loss bound $N-1$ that the cascade analysis builds on.","marker":"[15]"},{"why":"supplies the experimental and theoretical redundancy-parameter analysis of leakless strand displacement systems compared in Section 8.","marker":"[17]"},{"why":"gives the molecular-computation-at-equilibrium context and the AND gate implementation whose leak is bounded in Section 8.","marker":"[16]"}],"fun_headline_variants":["Athermic equilibrium solved: one algorithm sets every concentration","Entropy loss yields exponential DNA leak bounds","Algorithm tames off-target complexes in athermic systems","DNA logic leak reduced via equilibrium concentration bounds","One algorithm forces equilibrium concentrations in athermic chemistry"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The construction rests on $S$ being stable: every canonical reaction that introduces at least one off-target polymer must have $k(\\alpha)/l(\\alpha)>1$, and for the TBN applications this condition is checked only informally by arguing that particular reactions are worst-case.","fun_headline_variants_meta":{"raw":{"variants":["Athermic equilibrium solved: one algorithm sets every concentration","Entropy loss yields exponential DNA leak bounds","Algorithm tames off-target complexes in athermic systems","DNA logic leak reduced via equilibrium concentration bounds","One algorithm forces equilibrium concentrations in athermic chemistry"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00054,"raw_usage":{"total_tokens":2618,"prompt_tokens":1005,"completion_tokens":1613,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":621,"completion_tokens_details":{"reasoning_tokens":1539}},"tokens_in":621,"tokens_out":1613,"duration_ms":15510,"temperature":1.0,"reasoning_tokens":1539,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T16:44:40.367895+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the finite Hilbert basis of canonical reactions for a proposed stable set $S$ and test $k(\\alpha)/l(\\alpha)>1$ on every basis element with $l(\\alpha)>0$; any ratio $\\le 1$ refutes the conclusion that all off-target exponents exceed 1. For the translator cascade, solve the detailed-balance equations at the specified fuel and waste concentrations and check whether the leak output is below $c^{1/4+N/4}$ and whether more than half the signal reaches the output layer; failure of either inequality would falsify the worst-case-reaction claim.","supporting_citations":[{"cited_title":"Thermodynamic analysis of interacting nucleic acid strands","cited_arxiv_id":null,"evidence_quote":"supplies the free-energy function $g(x)$ whose minimizer, under mass conservation, defines equilibrium concentrations."},{"cited_title":"General mass action kinetics","cited_arxiv_id":null,"evidence_quote":"establishes that equilibrium minimizers satisfy detailed balance in mass-action systems, the condition the leveling algorithm enforces."},{"cited_title":"Foundations of chemical reaction network theory","cited_arxiv_id":null,"evidence_quote":"provides the chemical reaction network theory background on detailed-balance equilibria used to justify the free-energy formulation."},{"cited_title":"Rogers, David Soloveichik, Chris Thachuk, and Damien Woods","cited_arxiv_id":null,"evidence_quote":"introduces Thermodynamic Binding Networks and the entropy/stability notions the paper connects to concentrations."},{"cited_title":"Leakless DNA strand displacement systems","cited_arxiv_id":null,"evidence_quote":"defines the leakless DNA strand displacement translator family and the entropy-loss bound $N-1$ that the cascade analysis builds on."},{"cited_title":"Ellington, Erik Winfree, and David Soloveichik","cited_arxiv_id":null,"evidence_quote":"supplies the experimental and theoretical redundancy-parameter analysis of leakless strand displacement systems compared in Section 8."},{"cited_title":"Molecular computation at equilibrium via programmable entropy","cited_arxiv_id":null,"evidence_quote":"gives the molecular-computation-at-equilibrium context and the AND gate implementation whose leak is bounded in Section 8."}],"review_version":1}