{"id":"42c9b6b5-1e87-4210-bb36-a503635d50ed","arxiv_id":"2502.17461","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"If a reaction network has a weakly reversible deficiency zero realization for all rate constants, that realization is unique and can be found by an algorithm.","lead":"This paper proves a uniqueness theorem: if a reaction network can be given a stable 'weakly reversible deficiency zero' representation for every possible choice of reaction rates, then that stable representation is unique. The authors also supply an algorithm that finds this hidden stable network, which matters for identifying when complex biochemical models actually have simple, robust dynamics.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 4.4's proof is invalid: projecting ker Wκ1 rays does not yield ker Wκ0 rays, and the weak-reversibility contradiction fails because y0 is absent from Nκ0; however, Lemma 4.4 is repairable via Theorem 2.8, so Theorem 4.3 is likely correct but the written proof is incomplete.","rationale":"Agreeing with the reader: the weakest assumption is Lemma 4.4. I verified the proof gap in detail. The projection claim in Lemma 4.4 is false as stated (Wκ0 d = -c_m w_ym, not 0), and the weak-reversibility contradiction does not follow (y0 is not in Nκ0). However, the lemma is actually true by a one-paragraph argument using Theorem 2.8 and pointedness of WR0 linkage-class cones. Therefore the main theorem (uniqueness of WR0 realization independent of κ) is very likely correct, but the paper's proof is not rigorous as written. Independent support: Algorithm 2 is implemented in CoNtRol and the numerical checks pass, but that does not substitute for a proof. The result is a meaningful contribution and should be accepted after the proof of Lemma 4.4 is repaired. No formal verification is present.","tokens_in":13027,"tokens_out":13781,"duration_ms":104518,"concrete_test":"Write a corrected proof of Lemma 4.4 using Theorem 2.8 and Proposition 3.5: fix generic rate constants κ with no monomial cancellations; since N is WR0-realizable, (N,κ) has a WR0 realization Nκ and Theorem 2.8 implies Cone_N(y) ⊆ Cone_{Nκ}(y) for every source y of N. If 0 ∈ int(Cone_N(y)), then Cone_N(y) contains a line, so Cone_{Nκ}(y) contains a line, contradicting that Nκ's linkage classes are affinely independent (deficiency zero), which makes L_{Nκ}(y) a pointed simplicial cone. If this proof checks out, Theorem 4.3's remaining steps (Lemma 4.6 and Eq. (9)) are sound; if not, the theorem is unproved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Lemma 4.4 is load-bearing for Theorem 4.3 but its proof has two concrete errors. First, with Wκ1=[Wκ0|w_ym], a vector c=(d,c_m) in ker Wκ1 satisfies Wκ0 d = -c_m w_ym, not Wκ0 d=0. Thus the first m-1 coordinates of extreme rays of ker Wκ1 need not lie in ker Wκ0, and the claimed complete set of extreme rays of ker Wκ0 is not established; the support-partition argument fails for the unique ray whose support contains y0. Second, the contradiction asserts that Nκ0 is weakly reversible yet has no reactions yi→y0, which is not a contradiction because y0 has zero net vector in (N,κ0): the monomial x^{y0} is absent, so y0 is not a source complex (hence not a vertex) of Nκ0. Neither flaw is fatal to the lemma itself: choosing generic κ with no cancellations, y is a source complex of the WR0 realization Nκ, and Theorem 2.8 gives Cone_N(y) ⊆ Cone_{Nκ}(y); since Nκ is WR0, its linkage-class cone L_{Nκ}(y) is simplicial/pointed, so 0 cannot lie in int(Cone_N(y)). Thus Lemma 4.4 is true by a short alternative proof, but as written the paper's proof of Theorem 4.3 is incomplete and the central claim is CONDITIONAL on a repair.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies reaction networks N whose mass-action ODE systems admit a weakly reversible deficiency zero (WR0) realization for every choice of rate constants. The main result, Theorem 4.3, asserts that if such realizations exist for all κ, then the underlying WR0 network is independent of κ and unique. The authors propose Algorithm 2, which runs the known Algorithm 1 at κ=1 and then checks a cone-inclusion condition, to decide whether a given network is WR0-realizable and to return the unique WR0 network. Section 4 proves two lemmas (4.4 and 4.6) used to establish Theorem 4.3, and Section 5 reports an implementation in CoNtRol.","tokens_in":13307,"tokens_out":23896,"duration_ms":204242,"significance":"The uniqueness statement is natural and valuable: it turns the property 'has a WR0 realization for all rate constants' into a finite, checkable condition and strengthens the robustness conclusions available from the Deficiency Zero Theorem. The algorithmic component is a useful extension of Algorithm 1, and the paper is generally clearly written. The use of earlier published results by overlapping authors (Algorithm 1, Proposition 3.4, Theorem 3.6, Theorem 2.8) appears legitimate; these are used as tools, not as assumptions of the target conclusion. However, the proof of the key Lemma 4.4 is incomplete, so the main theorem is not yet established as written.","major_comments":[{"comment":"The proof of Lemma 4.4 is not valid as written. The step claiming that the first m−1 coordinates of the extreme rays of ker Wκ1 form a complete set of extreme rays of ker Wκ0 is false: for c=(d,c_m) in ker Wκ1, the equation Wκ1 c=0 gives Wκ0 d = −c_m w_{y_m}. If c_m>0, then d is not in ker Wκ0 at all, so the projected vectors are not extreme rays of ker Wκ0. Consequently the support-partition argument based on Proposition 3.4 does not apply. The final contradiction is also not a contradiction: because y0 has zero net vector in (N,κ0), y0 is not a source complex and not a vertex of Nκ0, so the absence of reactions yi→y0 does not violate weak reversibility of Nκ0. Since Lemma 4.4 is load-bearing for Theorem 4.3, this gap must be repaired; a correct proof will need to use the global 'for all κ' nature of the hypothesis rather than only the existence of a pointwise realization.","section":"Section 4, Lemma 4.4"}],"minor_comments":[{"comment":"In the sentence 'This implies that only the vertex y0 has zero net reaction vector in (N,κ)', the parameter should be κ0, not κ; likewise, the network denoted N0 near the end of the lemma should be Nκ0.","section":"Section 4, Lemma 4.4"},{"comment":"The claim 'there is a linkage class of Nκ1 with the same complexes as L and so c_l0 is an extreme ray of Nκ1 as well' is too terse; after equality in (8), counting forces exactly one extreme ray of Nκ1 whose support is L, and that ray must be a positive multiple of c_l0. Please spell out this counting argument.","section":"Section 4, Lemma 4.6"},{"comment":"The sentence beginning 'If, on the other hand, the algorithm goes through line 8' should be rephrased: failure of Cone_N(y)⊆Cone_N1(y) implies by Theorem 2.8 that N is not realizable by N1, and if N were WR0-realizable, Theorem 4.3 would force the realizing network to be N1, a contradiction. The current wording suggests that a particular κ0 with no N1-realization has been found, which is stronger than what the cone-inclusion check directly gives.","section":"Proof of Algorithm 2"},{"comment":"Please state explicitly whether reactions with y=y′ are allowed. If self-loops are permitted, Lemma 4.4 needs an additional hypothesis, since a source complex with only a self-loop has Cone_N(y)={0} and hence 0 lies in its relative interior.","section":"Section 2, Definition 2.1"},{"comment":"Equation (9) would benefit from a one-line justification: since wy lies in the relative interior of Cone_N(y), the face of L(y) generated by wy is the face generated by the whole cone Cone_N(y). This is true, but it is not immediate from the notation used.","section":"Section 4, Theorem 4.3 proof, Eq. (9)"}],"recommendation":"major_revision","confidential_remarks":"The main theorem is plausible and likely correct, but the proof of Lemma 4.4 is not a minor typo: both the projection argument and the weak-reversibility contradiction are invalid. The authors should be asked to replace that proof with a correct one. The algorithm's use of floating-point thresholds in Section 5 is a practical concern but not a mathematical one. I did not find circularity in the use of prior results by overlapping authors."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here is the short version: the main claim (Theorem 4.3) is a genuine extension of the fixed-parameter WR0 realization work of Craciun, Jin, and Yu, and Algorithm 2 is a natural, useful tool with a working CoNtRol implementation. But the proof as written has a gap in Lemma 4.4, which is load-bearing. The lemma is likely true and can be repaired, but the written argument does not do the job.\n\nThe paper does several things well. It states a clean question: if a reaction network is realizable by a weakly reversible deficiency zero network for every choice of rate constants, does the same WR0 network work for all of them? Theorem 4.3 says yes, and the uniqueness is not present in the earlier work. Algorithm 2 is correctly structured: run the existing fixed-parameter algorithm at κ=1, then check the cone inclusion condition from Theorem 2.8. The examples clarify where the fixed-parameter and all-parameters questions differ, and the software is a concrete deliverable.\n\nNow the soft spot. Lemma 4.4 claims that for a WR0-realizable network, no source complex has 0 in the interior of its cone, so the source monomial set is constant across rate constants. The proof has two errors. First, it projects extreme rays of ker Wκ1 onto the first m−1 coordinates and claims these are extreme rays of ker Wκ0. But for a vector c=(d,c_m) in ker Wκ1, we only get Wκ0 d = −c_m w_ym, not Wκ0 d=0. So the projection need not lie in ker Wκ0. Second, the contradiction with weak reversibility of Nκ0 does not follow, because y0 has zero net vector in (N,κ0) and is therefore not a source complex, hence not a vertex, of Nκ0. Absence of reactions into y0 is not a weak reversibility violation.\n\nBoth issues are repairable. As the stress-test note observes, choose a generic κ with no cancellations; then y is a source complex of the WR0 realization Nκ, and Theorem 2.8 gives Cone_N(y) ⊆ Cone_{Nκ}(y). Since Nκ is WR0 deficiency zero, its linkage class cone is pointed, so 0 cannot lie in the interior of Cone_N(y). That is a short alternative proof of Lemma 4.4. So Theorem 4.3 is probably correct, but the paper needs a corrected proof before it is fully trustworthy.\n\nWho should read it: people working on reaction network realizability, disguised toric systems, and computational structural analysis of mass-action networks. It is a solid subfield contribution, not a paradigm shift. I would recommend sending it to peer review, with a referee explicitly asked to check Lemma 4.4 and the projection argument.","headline":"A genuine extension of the WR0 realization program with a load-bearing but repairable gap in Lemma 4.4; worth sending to review.","tokens_in":13871,"tokens_out":2768,"would_cite":true,"duration_ms":22387,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92C42","37N25","52A20"],"pacs":[],"model":"deepseek-v4-flash","headline":"If a reaction network can be realized by a weakly reversible deficiency zero network for every rate vector, that realizing network is unique and rate-independent.","keywords":["weakly reversible networks","deficiency zero","reaction network realization","mass-action kinetics","rate constants","cone inclusion","dynamical equivalence","WR0-realizable"],"falsifier":"Take any network that passes Algorithm 2 and, for each source complex $y$, solve $\\sum_{y \\to y'} \\kappa_{y \\to y'}(y' - y) = 0$ with all $\\kappa_{y \\to y'} > 0$; if any solution exists, Lemma 4.4 is false and the uniqueness theorem loses its support.","tokens_in":12784,"feed_emoji":"🧪","tokens_out":11295,"duration_ms":89597,"temperature":0.7,"pith_summary":"This paper asks when a reaction network that is not itself weakly reversible and deficiency zero (WR0) can nevertheless produce, for every choice of rate constants $\\kappa$, the same mass-action dynamics as some WR0 network. The central result is Theorem 4.3: if such a WR0 realization exists for all $\\kappa$, then the realizing WR0 network is unique and does not change with $\\kappa$. The authors also give Algorithm 2, which certifies whether a given network is WR0-realizable and returns the unique realizing network by running a known fixed-rate algorithm at $\\kappa = 1$ and checking cone inclusions. A reader should care because WR0 networks have a unique, asymptotically stable equilibrium in every stoichiometric compatibility class, so a network that is merely WR0-realizable inherits these stability guarantees no matter what its actual rate constants are.","feed_headline":"A network's stable disguise is fixed, no matter the rates","feed_subtitle":"Find one weakly reversible, deficiency-zero network and stability is certified for every rate constant.","key_machinery":"The argument is carried by the cone geometry of a mass-action system. For a source complex $y$, $\\operatorname{Cone}_{\\mathcal{N}}(y)$ is the cone generated by the reaction vectors $y' - y$ for reactions $y \\to y'$ leaving $y$, and $L_{\\mathcal{N}}(y)$ is the larger cone generated by all complexes in $y$'s linkage class inside a WR0 realization. The net reaction vector $w_y$ is the coefficient column of the monomial $x^y$ after grouping terms. The proof uses Algorithm 1, which finds the unique WR0 realization of a fixed mass-action system from the extreme rays of $\\ker W \\cap \\mathbb{R}_{\\ge 0}^m$; Lemma 4.4 is meant to ensure no source monomial can vanish; Lemma 4.6 shows the linkage class partition is rate-independent; and Theorem 3.6 supplies uniqueness of the WR0 realization for a fixed ODE system.","core_discovery":"The paper's central claim is that the relation '$\\mathcal{N}$ is realizable by $\\mathcal{N}'$ with $\\mathcal{N}'$ weakly reversible and deficiency zero for every $\\kappa$' is rigid: the network $\\mathcal{N}'$ is independent of $\\kappa$. More precisely, if $\\mathcal{N}$ is WR0-realizable, then there exists a unique WR0 network $\\mathcal{N}'$ such that for any rate constants $\\kappa$ there are rate constants $k_\\kappa$ for $\\mathcal{N}'$ with $f_{\\mathcal{N},\\kappa} = f_{\\mathcal{N}',k_\\kappa}$. The proof proceeds by showing that the set of source complexes is rate-independent, that the linkage class partition of the realizing network is rate-independent, and that the face of the linkage-class cone generated by each net reaction vector is rate-independent; these three facts force the reaction set of $\\mathcal{N}'$ to be fixed.","pith_inferences":["If Lemma 4.4 fails, the uniqueness theorem could fail too: rate constants might change which source complexes appear, and the 'unique' realization could depend on the parameter vector.","The set of rate constants for which a given network has a WR0 realization (Question 3 of the paper) is plausibly described by polynomial inequalities defining cones; this is a concrete next problem the paper leaves open.","The numerical tolerances used in the implementation suggest that an exact symbolic computation of extreme rays would make the certificate reliable near cone boundaries, where small numerical errors could flip the output."],"forward_implications":["Because the realizing WR0 network is unique, checking WR0-realizability needs only one run of the fixed-rate algorithm (at all rate constants equal to 1) followed by cone-inclusion checks.","Every mass-action system generated by a WR0-realizable network inherits the WR0 stability guarantees — a unique equilibrium in each compatibility class, asymptotically stable — for any rate constants.","The set of source complexes of the original network is rate-independent, so no source monomial can disappear from the dynamics as parameters vary.","The unique realizing network $\\mathcal{N}'$ provides a fixed certificate: once found, it can be used to decide dynamical realizability by WR0 networks for the whole network, not just for one rate vector."],"supporting_citations":[{"why":"Supplies Algorithm 1, the fixed-rate WR0 realization routine that the new Algorithm 2 runs with all rates equal to 1.","marker":"[33]"},{"why":"Published version of the fixed-rate algorithm that answers Question 1 and is the main technical tool behind the proof.","marker":"[51]"},{"why":"Proves Theorem 3.6, that a mass-action system has at most one WR0 realization, which the new uniqueness result extends from one ODE system to the whole family generated by a network.","marker":"[40]"},{"why":"Gives the cone-inclusion characterization of dynamical realizability (Theorem 2.8) that Algorithm 2 checks.","marker":"[50]"},{"why":"Establishes that WR0 networks are complex balanced with asymptotically stable equilibria, the dynamical property that makes WR0 realizations worth detecting.","marker":"[9]"},{"why":"Provides the deficiency-zero theory, including affine independence of complexes in a linkage class, used in Proposition 3.5 and in the proof of Lemma 4.6.","marker":"[13]"}],"fun_headline_variants":["One stable network fits every rate constant","Unique rate-proof network structure unveiled","Algorithm finds the one network that beats all rates","Deficiency-zero shape that ignores rate variations"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof in Section 4 depends on Lemma 4.4 — no reactant complex can ever have its outgoing reaction vectors sum to exactly zero — and as written that lemma is not established, because the zero-sum complex it constructs is not a complex of the network used to reach the contradiction.","fun_headline_variants_meta":{"raw":{"variants":["One stable network fits every rate constant","Unique rate-proof network structure unveiled","Algorithm finds the one network that beats all rates","Deficiency-zero shape that ignores rate variations"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000133,"raw_usage":{"total_tokens":1050,"prompt_tokens":776,"completion_tokens":274,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":392,"completion_tokens_details":{"reasoning_tokens":220}},"tokens_in":392,"tokens_out":274,"duration_ms":3952,"temperature":1.0,"reasoning_tokens":220,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T15:19:06.211923+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take any network that passes Algorithm 2 and, for each source complex $y$, solve $\\sum_{y \\to y'} \\kappa_{y \\to y'}(y' - y) = 0$ with all $\\kappa_{y \\to y'} > 0$; if any solution exists, Lemma 4.4 is false and the uniqueness theorem loses its support.","supporting_citations":[{"cited_title":"An algorithm for finding weakly reversible deficiency zero realizations of polynomial dynamical systems","cited_arxiv_id":"2205.14267","evidence_quote":"Supplies Algorithm 1, the fixed-rate WR0 realization routine that the new Algorithm 2 runs with all rates equal to 1."},{"cited_title":"Craciun, J","cited_arxiv_id":null,"evidence_quote":"Published version of the fixed-rate algorithm that answers Question 1 and is the main technical tool behind the proof."},{"cited_title":"Craciun, J","cited_arxiv_id":null,"evidence_quote":"Proves Theorem 3.6, that a mass-action system has at most one WR0 realization, which the new uniqueness result extends from one ODE system to the whole family generated by a network."},{"cited_title":"Craciun and C","cited_arxiv_id":null,"evidence_quote":"Gives the cone-inclusion characterization of dynamical realizability (Theorem 2.8) that Algorithm 2 checks."},{"cited_title":"Horn and R","cited_arxiv_id":null,"evidence_quote":"Establishes that WR0 networks are complex balanced with asymptotically stable equilibria, the dynamical property that makes WR0 realizations worth detecting."},{"cited_title":"Feinberg","cited_arxiv_id":null,"evidence_quote":"Provides the deficiency-zero theory, including affine independence of complexes in a linkage class, used in Proposition 3.5 and in the proof of Lemma 4.6."}],"review_version":1}