Reviewer3 (AI review)

September 28, 2026

AI Review (reviewer3.com)

Comments (5)

1. Number of highest-scoring bridges — Editorial

, with the full ranked scan provided as bridge_scan.csv, the two highest - scoring bridges are both anchored on Lys21 (Table 3):

Section 3.6 states that "the two highest-scoring bridges are both anchored on Lys21 (Table 3)". However, Table 3 lists three bridges with the maximum score of 8.0, and the third one (Ser25–Ser29) is not anchored on Lys21.

2. Tail residue burial contradiction — Editorial

ute a second, smaller patch. In contrast Lys21, Glu24, Ser25, Gly28, Ser29, Ser32 and the rest of the tail lose no accessible surface at all on binding (ΔSASA = 0 Ų).

Section 3.5 states that "the rest of the tail lose no accessible surface at all on binding (ΔSASA = 0 Ų)." However, Supplementary Table S1 shows that tail residues Ser30 and Arg34 have non-zero buried surface area on binding (9.3 Ų and 29.4 Ų respectively for the native peptide, and 16.3 Ų and 31.6 Ų respectively for the cyclised peptide HSP-3).

3. Lennard-Jones repulsion prevents proper ring closure — Major

LINK record. The achieved Nζ – Cγ distance, 1.82 Å, is reported as - is and is the principal geometric approximation in this work.

The Lys21–Asp25 lactam was closed using a harmonic distance restraint of 1.33 Å but only achieved a distance of 1.82 Å because the bond was enforced geometrically rather than parameterized. However, if the covalent bond was not explicitly defined in the topology file, the force field would still calculate non-bonded Lennard-Jones (van der Waals) repulsion between the Lys21 Nζ and Asp25 Cγ atoms. At a distance of 1.82 Å, this repulsion is extremely strong and directly opposes the harmonic restraint, preventing proper closure and potentially introducing artificial strain into the local backbone. To resolve this, these two atoms should be excluded from non-bonded interactions in the topology file (or a standard covalent bond with appropriate parameters should be defined) and the minimization and relaxation should be re-run.

4. Lack of relaxation in transplanted complexes — Major

de residue. No force field was applied to the transplanted complexes and no interaction energy is reported: the comparison is contact - level. The

The binding footprints of HSP-3 in VPAC1 and VPAC2 are compared by superimposing the peptide into the static receptor structures without performing any energy minimization or relaxation of the complex. Because the receptors and templates have different side chains and conformations, simple superposition is highly likely to introduce severe steric clashes between the peptide and the receptor. These unresolved clashes can artificially distort the solvent-accessible surface area (SASA) and contact calculations, making the conclusion of near-identical global interface sizes and localized differences unreliable. A brief energy minimization of both complexes using a standard force field should be performed to resolve clashes before calculating the interface metrics.

5. Inappropriate template for proline-rich C-terminal tail — Major

t conformation in VPAC2. The template differs from helospectin at 24 of 38 positions, and the C - terminal extension is taken from a micelle - bound NMR ensemble of a different peptide rather than from any receptor complex. An independent prediction of the hel

The C-terminal tail of helospectin I (residues 28–38, sequence GSSTSPRPPSS) was threaded directly onto the template of PACAP-38 (sequence GKRYKQRVKNK) from PDB 2D2P. These two sequences have completely different physicochemical properties and secondary structure propensities; helospectin I is highly proline-rich and polar, whereas PACAP-38 is highly basic. Proline residues have extremely restricted backbone dihedral angles and high barriers to conformational change. A brief 100 ps relaxation is entirely insufficient to overcome these barriers and find the correct conformation of a proline-rich tail. Consequently, the modeled conformation of the tail (including Gly28 and Ser29) is highly unreliable, making any conclusions about their solvent accessibility or lack of receptor interaction suspect. To address this, de novo folding methods such as AlphaFold3 or ColabFold should be used to model the tail conformation, or the claims regarding these residues should be tempered.

xPeerd (AI review)

September 28, 2026

AI Review (xPeerd — DAReview)

Summary

This manuscript presents a computational peptide-design study that selects helospectin I from the Heloderma venom proteome as a scaffold for VPAC-mediated genital vasodilation and proposes a macrocyclised analogue, HSP-3, intended for erectile dysfunction and female sexual arousal disorder. The core workflow includes proteome triage, dual-template receptor-bound modelling, interface mapping through ΔSASA, geometric bridge scanning, explicit construction of a Lys21–Asp25 lactam, short implicit-solvent stress MD, and a contact-level VPAC1/VPAC2 comparison. The main strength is conceptual coherence: the author clearly frames a medicinal-chemistry hypothesis around a specific receptor and a specific structural constraint, and the manuscript is unusually transparent about several limitations. The strongest result is not experimental efficacy but a structure-based proposal that residues around 21–25 may be modifiable without directly disrupting the modeled VPAC2 interface.

The main weakness is that nearly every downstream conclusion depends on a single unvalidated threading assumption, using a template that differs at 24 of 38 positions and a tail borrowed from a different peptide in a micelle-bound NMR ensemble. The dynamics, selectivity comparison, and macrocycle assessment are all too limited to support strong claims beyond hypothesis generation. The work is therefore potentially interesting as a design note or computational hypothesis paper, but not yet as persuasive evidence for lead nomination or translational prioritization.

Potential Major Revisions

1. The manuscript’s central structural premise is insufficiently validated for the weight placed upon it.

The paper explicitly states that “The receptor-bound backbone of the model is inherited from the template by construction” and that the restrained agreement to template “is a property of the protocol and not evidence that helospectin I adopts that conformation in VPAC2” in Section 3.4. It also acknowledges that “The template differs from helospectin at 24 of 38 positions” and that “An independent prediction of the helospectin–VPAC2 complex, by AlphaFold2-Multimer or a co-folding model, is the calculation that would support or break this assumption, and it was not run here.”

This is the manuscript’s most important vulnerability. The interface map, free-face identification, bridge scan, macrocycle placement, and receptor selectivity discussion all inherit this uncertainty. In its current form, the paper sometimes reads as if the geometric findings have greater evidentiary force than the model justifies. Before submission, the manuscript needs at minimum one independent structural cross-check of the helospectin–VPAC2 complex. Since the author himself identifies this as “the single most valuable addition,” omitting it is difficult to justify.

2. The lead-selection logic is based on geometric plausibility rather than binding or function, and the manuscript should narrow its claims accordingly.

The paper states in Section 6 that “No binding affinity, no cAMP response, no tissue or animal experiment is reported here. The analogue series is a hypothesis to be tested, not a result.” It further concedes “No affinity or free-energy estimate,” “No energy calculation of any kind,” and that “all rankings here are structural.”

That candor is welcome, but the conclusion still proposes “HSP-3, [Asp25] helospectin I-NH₂ cyclised between Lys21 and Asp25, as the lead” and frames the analogue as a lead for local genital vasodilation. The word “lead” is too strong unless supported by at least orthogonal computational evidence, such as co-folding validation, relaxed complex comparison, or approximate binding-energy ranking. At present, the analogue is better described as a prioritized design hypothesis rather than a lead candidate. The claims should be softened consistently throughout the abstract, discussion, and conclusions.

3. The macrocycle construction method is physically approximate and may bias the positive interpretation of conformational retention.

Section 2.6 states that “The lactam is enforced geometrically rather than by an explicit force-field template,” because “no amber parameter set exists for a Lys–Asp isopeptide.” The authors report “The achieved Nζ–Cγ distance, 1.82 Å, is reported as-is and is the principal geometric approximation in this work.” Section 6 repeats that “The lactam is geometric, not parameterised.”

This is not a minor detail. If the covalent linkage is represented by a restraint rather than a proper bonded topology with angle and torsion terms, then the observed retention of helix and small RMSD shift may partly reflect construction choices rather than a chemically realistic macrocycle. The manuscript should either provide a parameterization route, a QM-derived surrogate template, or at least a much more explicit methodological sensitivity analysis showing that the central design conclusion is robust to the approximation. Without that, the key positive structural result remains provisional.

4. The molecular dynamics evidence is too weak to support meaningful stability claims.

The Methods report “A single trajectory was run per system” for 300 ps at 350 K in implicit solvent, with framewise dispersions rather than replicate uncertainty estimates. The limitations section correctly states these simulations “do not establish thermodynamic stability” and that “no replicate simulations were run.”

Given this, the sentence in the abstract and results that the cyclised peptide “retains more helical hydrogen bonding than the native peptide” should be presented more cautiously, because the evidence is only a brief stress test under a high-temperature implicit-solvent protocol. The manuscript would be substantially stronger with multiple independent trajectories, longer timescales, explicit solvent, and a clearer rationale for why 350 K stress testing is informative for this system. In the current form, these data support only a tentative suggestion of altered short-timescale conformational persistence, not stability.

5. The selectivity discussion risks overinterpretation relative to the actual analysis performed.

The VPAC1/VPAC2 comparison is described as a “contact-level” analysis of transplanted static poses, with “No force field” and “no interaction energy” applied. The limitations further state that “Selectivity is compared, not predicted” and that the expectation of VPAC2 preference “still rests substantially on the analogy with Ro 25-1553.”

That limitation is serious because selectivity is presented as a key safety lever in the discussion: “The VPAC1/VPAC2 split is the main pharmacological handle available for separating genital from systemic effects.” Yet the manuscript does not establish that HSP-3 actually improves that split. Any phrasing that implies emergent VPAC2 selectivity of HSP-3 should be trimmed unless supported by stronger calculations or experimental functional assays. The paper is strongest when it presents selectivity as an open question.

6. The translational framing sometimes outpaces the evidence and should be tightened.

The manuscript proposes a preclinical cascade and argues that a local VPAC agonist may address cases in which PDE5 inhibitors are blunted, including diabetic endothelial dysfunction and post-surgical or post-radiation injury. This is mechanistically plausible, but the chain from receptor engagement model to genital tissue efficacy is still indirect. The paper itself admits “Whether VPAC2 selectivity is in fact the correct direction for genital tissue is a further open question” best resolved by receptor-expression profiling and cAMP assays.

The clinical indication sections should therefore be written more clearly as rationale, not inferred applicability. Otherwise, the paper risks appearing to overproject from structural modeling into therapeutic positioning.

7. Internal integrity check: no clear fabrication signal, but there are some paper-mill-adjacent features that warrant correction.

I do not see strong evidence of data fabrication in the manuscript. In fact, the unusually explicit limitation statements argue against a typical paper-mill style. However, there are integrity-adjacent issues: the author contributions section appears inconsistent, because the manuscript begins with a single named author, yet later states “Both authors read and approved the final version,” with a broken sentence fragment before it. Also, the acknowledgments disclose use of “AI claude in science ) for protein to assist with scripting,” which is grammatically malformed and insufficiently specified. These are not proof of misconduct, but they create avoidable credibility problems and should be corrected before review.

Potential Minor Revisions

1. Typographic and grammatical issues

The manuscript contains multiple visible language and formatting problems that should be corrected systematically:

• The front matter repeatedly runs section labels into body text, for example “Preprint not peer reviewedResults” rather than a clean section break. This appears in the early pages around the abstract/results transition.
• “This preprint research paper has not been peer reviewed” and SSRN boilerplate repeatedly interrupt paragraphs and page flow, harming readability. These should be removed or confined to headers/footers in the final version.
• In the introduction, the sentence “Men with diabetes have about 3.6-fold higher odds of erectile dysfunction than healthy controls, and review literature indicates that ED tends to present 10–15 years earlier in diabetic men.(2).” contains incorrect punctuation around the citation: “men.(2).”
• The methods contain inconsistent hyphenation and dash use, such as “co-folding” versus “co folding” risk in surrounding text and mixed dash typography around residue ranges and receptor complexes. A line-by-line normalization is needed.
• “selectivity_summary. json” contains an erroneous space before “json.”
• The author contributions statement is grammatically broken: “A.K.K. conceived the study, designed and performed the computational analyses, prepared the figures and tables, and drafted the manuscript. interpretation of the results and to critical revision of the manuscript for important intellectual content. Both authors read and approved…” This is incomplete and inconsistent.
• The acknowledgments sentence “The authors used AI claude in science ) for protein to assist with scripting of the structural analyses” is malformed and should be rewritten precisely.
• Declarations use informal wording such as “Funding: NO.” and “Conflicts of interest. NO .” These should be replaced with standard academic phrasing, e.g., “The author reports no funding” and “The author declares no conflicts of interest.”
• There are punctuation anomalies in the references, for example “PMID: 1313797..” with a double period.
• Several instances of “VPAC1R–VIP–Gs” and similar notations should be checked for consistent capitalization and typography of receptor names, especially “Gs” versus “Gs protein” style.

2. Formatting issues

• The manuscript contains intrusive SSRN/preprint boilerplate in the middle of paragraphs and section transitions, which should be removed in the review copy or camera-ready form.
• Tables in the extracted text appear underformatted and may require clearer column headings, alignment, and species/UniProt presentation. Table 1 especially should distinguish peptide names, accession numbers, biological activity, and decision rationale more cleanly.
• Some figure legends are informative, but the manuscript would benefit from explicit statement in each legend about whether the displayed structure is restrained, transplanted, or derived from the unvalidated model. This is especially important for Figures 3 and 7.
• Abbreviation density is high. A short abbreviation list or first-use expansion audit would improve readability for interdisciplinary reviewers.

3. Conceptual precision and wording refinements

• The phrase “answers that call computationally” in the introduction is rhetorically confident relative to the actual scope. A more cautious wording would better match the evidence.
• The discussion sentence “Few peptide-engineering campaigns start that close to the target” is too promotional for a manuscript whose structure is not independently validated.
• The paper should distinguish consistently between “VPAC-mediated vasodilation,” “VPAC2-biased design hypothesis,” and “local genital efficacy.” These are not equivalent evidentiary levels.

4. AI content analysis for this post-2021 manuscript

Estimated percentage of AI-generated or AI-assisted prose: moderate, approximately 20–40%.

Basis for this estimate:

• The manuscript discloses AI assistance explicitly in the acknowledgments: “The authors used AI claude in science ) for protein to assist with scripting of the structural analyses, with drafting, and with language editing.”
• The prose is highly uniform, polished, and rhetorically balanced, with repeated meta-discursive constructions such as “This is the structural result on which the rest of the paper depends” and “This is the single most valuable addition before submission,” which are common in AI-assisted revision.
• The limitations section is unusually exhaustive and formulaically organized, which may reflect AI-supported drafting.

Highlighted likely AI-assisted sections:

• The limitations section is the clearest candidate for AI-assisted drafting because of its repetitive, standardized cadence: “No experimental data,” “Homology model, not a structure,” “No AlphaFold2 or co-folding cross-check,” “No affinity or free-energy estimate,” “The lactam is geometric, not parameterised,” “Dynamics are short and unreplicated,” “Selectivity is compared, not predicted.”
• Portions of the discussion also show AI-like summarizing and contrastive phrasing, for example “Why the interface analysis matters more than sequence intuition” and “Product logic: on-demand versus chronic.”

Assessed epistemic impact:

• Low to moderate if the AI role was limited to drafting, editing, and scripting assistance as disclosed.
• Moderate to high if AI also influenced unverified scientific inference without adequate human validation.
• In this case, the main epistemic risk is not AI use itself, but whether AI-assisted writing may have made speculative claims sound more decisive than the underlying computational evidence warrants. Because the author also includes many caveats, the net impact is manageable, but stronger claim discipline is needed.

Recommendations

1. Add an independent structural validation step before submission. Run AlphaFold2-Multimer or another credible peptide–receptor co-folding workflow for helospectin I with VPAC2, and compare residue-level interface agreement with the current threaded model. This single addition would materially improve the manuscript’s credibility because it directly tests the foundational assumption already acknowledged in Section 3.4.

2. Recast HSP-3 as a prioritized hypothesis, not a validated lead. Unless new binding, signaling, or orthogonal structural evidence is added, replace “lead” terminology with “proposed analogue,” “priority design,” or “testable candidate.” This would align the abstract and conclusion with the actual evidence base.

3. Strengthen the treatment of the covalent linkage. Either parameterize the Lys–Asp isopeptide properly, import a chemically credible surrogate, or provide a sensitivity analysis demonstrating that the chosen bridge ranking is stable to modelling assumptions. As written, the geometric restraint approach is acceptable for exploratory prototyping, but not for strong structural inference.

4. Expand and replicate dynamics if stability is to remain a claimed advantage. Use longer simulations, multiple seeds, and ideally explicit solvent. If computational resources are limited, reduce the interpretive weight of the current MD and state more plainly that it is only a qualitative stress test.

5. Remove any implication that VPAC2 selectivity has been shown. Keep the manuscript’s strongest version: it identifies a plausible structural route to test selectivity, but does not yet predict or demonstrate it convincingly.

6. Clean the manuscript’s authorship, declaration, and AI-disclosure sections. Resolve the one-author/two-author inconsistency, rewrite the malformed AI acknowledgment, and standardize funding/conflict statements to accepted journal style. These are small repairs with disproportionate impact on trustworthiness.

7. Consider positioning the work as a hypothesis-generating computational design study. In its current state, the manuscript may be better received if framed explicitly as a design and prioritization paper, not a near-translational lead-optimization study. That framing is already supported by the manuscript’s own best sentence: “the work is offered as a design hypothesis for experimental testing, not as evidence of clinical utility.”

XSci (AI review)

September 28, 2026

Referee Report

Manuscript: Structure-guided macrocyclization of the Gila monster venom peptide helospectin I: a computational design study of VPAC-targeted analogues for vasculogenic erectile dysfunction and female sexual arousal disorder

Recommendation: Major revision

Summary and Recommendation

This computational study asks whether a peptide from Heloderma venom can be re-engineered into a conformationally constrained VPAC agonist for local, on-demand genital vasodilation, on the rationale that Gs–adenylate cyclase–cAMP signalling relaxes cavernosal smooth muscle by a route that runs in parallel to NO–cGMP and should therefore survive where PDE5 inhibition is blunted. The authors triage the reviewed Heloderma proteome by annotated target and toxicity to helospectin I, build a receptor-bound model on a dual template combining PACAP-27 as bound to human VPAC2 (PDB 7VQX) with the micelle-bound PACAP-38 NMR ensemble (PDB 2D2P), compute a per-residue burial-on-binding profile, and scan every i,i+3, i,i+4 and i,i+7 pair in the helical segment against geometric, accessibility, interface and conservation criteria. Two lactams score highest, both anchored at the native Lys21: an i,i+3 ring to Glu24 requiring no sequence change, and an i,i+4 ring to a substituted Asp25, designated HSP-3 and nominated as the lead. The cyclised peptide is compared with the native under thermal stress and transplanted into VPAC1 for a contact-level receptor comparison, and a nine-member analogue series and six-step preclinical cascade are proposed.

The manuscript has real merits. Its central observation, that a naturally occurring venom peptide already presents several features that had to be engineered synthetically into VIP to obtain a long-acting subtype-preferring agonist, is genuinely novel and is checked correctly against the VIP sequence. The decision to place a constraint on a solvent-facing face identified structurally, rather than on positions chosen by sequence intuition, is the right methodological instinct, and the contrast drawn between the Lys21 case and the Glu9–Lys12 case illustrates it well. The authors report their own limitations more frankly than is customary, including the circularity of the restrained model, the geometric rather than covalent nature of the lactam, and the fact that the reported dispersions are frame-to-frame standard deviations rather than errors over replicates. They also report the RMSD metric that does not favour their design. Against this, three sets of problems currently prevent publication: the quantitative results are each compromised by a feature of the protocol rather than by any conceptual error; the pharmacological premise and the selectivity rationale are asserted from database annotation where primary literature exists and, in two places, where published experimental work runs the other way; and the manuscript's own internal record is inconsistent to the point where a reader cannot verify the design. Major revision is required, but almost everything needed is within reach of the authors' existing methods.

Major Concern Category 1: Validity of the Structural Model and Its Simulations

1.1 The footprint that the design rests on is a property of the restraint protocol and is reported without any sensitivity analysis

Section 3.4 is candid that the Cα atoms of residues 1–27 were restrained to the PACAP-27 pose throughout relaxation, so that the 0.06 Å agreement with the template is a property of the protocol rather than evidence about helospectin. What the manuscript does not follow through is that this dependence does not stop at the RMSD; it propagates directly into Section 3.5, which the authors themselves call the structural result on which the rest of the paper depends. The ΔSASA values are computed from a single static pose whose backbone was held at a template derived from a peptide differing at 24 of 38 positions, with side chains relaxed from what appears to be a single rotamer set and no systematic repacking. The assignments that matter most are the most sensitive to this: Lys21, Glu24 and Ser25 are reported as burying exactly zero, and a hard zero for a long, flexible lysine side chain derived from one energy-minimised snapshot conveys a precision the calculation cannot support. The same concern applies to the 15 Ų interface threshold used in the bridge scan, which is applied to point values carrying no stated uncertainty. Repeating the analysis on an independently predicted complex, computing burial over a short restrained-backbone ensemble so that each value carries a dispersion, and demonstrating that the interface classification is stable across a range of thresholds would convert a protocol artefact into a finding.

1.2 The macrocycle is a harmonic restraint present in only one arm of the dynamics comparison, and the controlling simulation is missing

Section 2.6 describes the lactam as enforced geometrically, closing to 1.82 Å against an amide reference of 1.33 Å, and Section 3.7 states that the ring is therefore not covalently closed in the calculation. Section 2.7 then states that this harmonic bond was retained in the production dynamics. The comparison in Section 3.8 is consequently not between a macrocyclic and a linear peptide under identical Hamiltonians, but between a peptide carrying an additional non-physical potential that pulls residues 21 and 25 together and one that does not. Unwinding of a helical turn is precisely the motion such a restraint resists, so the reported excess of roughly three i→i+4 hydrogen bonds is at least partly mechanical, and the fact that the restraint never reaches its equilibrium value means it exerts force throughout. Compounding this, HSP-3 differs from the native peptide in two ways, the S25D substitution and the ring, and no linear [Asp25]helospectin I was simulated, so the observed difference cannot be attributed to cyclization rather than to the substitution. Simulating that third system, parameterising the isopeptide bond properly or demonstrating that a sham restraint on the native peptide does not reproduce the effect, and reporting the restraint energy over time would together resolve this.

1.3 The simulation and receptor-comparison evidence is under-powered, internally discordant, and confounded by an avoidable template mismatch

Single 300 ps implicit-solvent trajectories at 350 K cannot distinguish a systematic difference from one stochastic unwinding event, and the two reported metrics disagree in the window that carries the claim: over the final 50 ps the hydrogen-bond count favours HSP-3 while the backbone Cα RMSD reverses and favours the native peptide, 5.04 ± 0.27 Å against 4.50 ± 1.14 Å. The authors explain this as tail-dominated, which is plausible but untested; computing RMSD over residues 2–27 alone would test it directly. Replicate trajectories reporting between-replicate standard errors would settle the rest. The receptor comparison in Section 3.9 suffers an analogous problem. Its first acknowledged caveat, that the two templates were solved with different agonists, is not merely a caveat but the whole result, since the only positive finding is a set of local differences of 16 to 66 Ų that the caveat concedes may report on the template. The confound is removable: the study cited as reference 30 also produced a VPAC1R–PACAP27–Gs complex, and an independent earlier study reported another, either of which would isolate receptor identity as the sole variable. Neither is mentioned. Nor is any noise floor established for per-residue burial differences obtained from a rigid transplant with no repacking, no rotamer sampling and no sensitivity analysis on the superposition range or the contact cut-offs.

Major Concern Category 2: The Pharmacological Premise and the Attribution of Selectivity

2.1 The premise that helospectin's vasodilation is VPAC-mediated is not established, and contrary published evidence is not engaged

Everything in the manuscript follows from the proposition that helospectins are VPAC agonists that relax vascular smooth muscle through Gs–cAMP–PKA signalling. The support offered is UniProt annotation, a 1967 report of the physiological effects of whole venom, and a 1983 study of venom interaction with VIP receptors in intestinal epithelium. None is a demonstration that purified helospectin relaxes vascular smooth muscle through VPAC receptors. The question has been asked experimentally: in the intact hamster cheek-pouch microcirculation, helospectin I and II each evoked potent vasodilation that was unaffected by VIP10–28, by PACAP6–38 or by indomethacin, and the authors of that study concluded the response was not transduced by VIP or PACAP receptors. That work is not cited. It does not settle the mechanism in human cavernosal tissue, and the antagonists used have known limitations, but it is the most direct published test of the present premise and cannot be passed over. The revision should engage it explicitly, and the assay cascade should be amended so that receptor mediation is interrogated rather than assumed: step 3 currently tests only whether relaxation survives NO-synthase blockade, which a non-receptor-mediated effect would also do.

2.2 Candidate selection rests on curated annotation where primary comparative pharmacology exists

Section 3.1 states that helospectin I was selected partly because it is annotated with the highest affinity for human VPAC2. An annotation is not a measurement, and the primary pharmacology of this peptide family has been studied: comparative work on recombinant rat and human VIP1 and VIP2 receptors reported that helodermin and helospectin are selective for the human VPAC2 subtype, while emphasising that these analogues discriminate between rat and human receptors. Citing the measurement would strengthen the selection and simultaneously surface a risk the manuscript does not acknowledge, namely that the rat efficacy models proposed in cascade steps 4 and 5 may not report on human pharmacology. Two further selection questions deserve a sentence each. The exclusion of gilatoxin, phospholipase A2 and helothermine as toxic is reasonable but the threshold is unstated, and toxicity of whole venom is a different property from unsuitability as a scaffold. And the preference for the longest member of the group is never justified against the countervailing considerations of synthesis cost and mucosal delivery that the manuscript raises elsewhere, when helodermin at 35 residues retains 89% identity over the core.

2.3 The VPAC2-selectivity rationale is misattributed, and the objective function itself is of conceded unknown sign

The Ro 25-1553 analogy is the manuscript's most persuasive move, and it is built on a premise that has been tested and found otherwise. The published dissection of that compound reports that selectivity for the human VPAC2 receptor is supported mainly by N-terminal acetylation, by the lysine at position 12 and by the C-terminal extension, whereas the 21–25 lactam contributed to affinity but participated only marginally in selectivity; a selective VPAC2 antagonist lacking the lactam entirely was described in the same work. This is not cited. Three consequences follow. Table 4 should not list VPAC2 preference among HSP-3's expected gains. The element of the reference compound that helospectin does not natively carry is the N-terminal acetyl group, which the manuscript assigns to HSP-2 and flags as risking class B efficacy, and that tension is the interesting unsolved problem in this series. And the case for HSP-3 over the zero-substitution HSP-3b weakens, since a ring contributing affinity rather than selectivity makes the cheaper option more attractive by the paper's own economy criterion. Standing behind all of this is the Discussion's own concession that whether VPAC2 selectivity is the correct direction for genital tissue remains open, which, if taken seriously, means the design is optimising a quantity whose sign is unknown and that the tenfold-selectivity go/no-go could reject an effective compound. That concession belongs in the premises, not in a subordinate clause.

Major Concern Category 3: Internal Consistency of the Record and the Design Built Upon It

3.1 The primary sequence is inconsistent, and the residue numbering that carries every design claim therefore cannot be verified

Section 2.1 states that helospectin I comprises 38 residues and gives a sequence containing 37. Table 4 gives a 38-residue sequence differing by an additional proline in the C-terminal segment, and Supplementary Table S1 assigns positions 32 to 38 consistently with Table 4 and inconsistently with the Methods. The Methods and the Tables therefore describe two different peptides. This is consequential rather than clerical, because every substantive claim is expressed in residue numbers: the Lys21 ring anchor, the S25D substitution defining the lead, the Ser32 glycosite proposed for HSP-7, the S30K substitution underlying HSP-6, and the entire burial profile. If the tail carries one fewer residue, every position from 33 onward shifts. The manuscript offers no way to adjudicate, because the primary sequence determination is cited only as historical background rather than as the authority for the sequence, and because two UniProt accessions from two species are treated as interchangeable without either sequence being shown. The revision should state the sequence once, authoritatively, with accession and version, identify the species modelled, and reconcile the Methods with both tables.

3.2 The central footprint claim is contradicted by the manuscript's own supplementary data, and the table that selects the lead cannot be reproduced

Section 3.5 and the Abstract both state that the named solvent-facing residues and the C-terminal tail lose no accessible surface on binding. Supplementary Table S1 reports Ser30 burying 9.3 Ų and Arg34 burying 29.4 Ų, both within the region covered by that statement and the latter above the manuscript's own 15 Ų interface threshold. Section 3.9 then reports increased burial at Arg34 and Ser30 in the VPAC1 pose, an analysis coherent only if those residues bury a non-zero amount at VPAC2. Supplementary Table S1 additionally carries two burial columns giving different values for the same residue in the same complex, differing by more than 20% at Lys12, with the Methods explaining neither and the text drawing on both, which also falsifies the claim in Section 2.8 that the implementation agrees with reference values to within 5 Ų per residue. Table 3 compounds this. It is typographically fractured, its mutations column is misaligned so that the top-ranked pair appears to require no substitution, it assigns the 25/29 pair a lactam substitution while the text describes a hydrocarbon staple there, it describes a three-way tie at a score of 8.0 as a two-way result, and its composite scores cannot be recomputed because the weights converting four pass/fail criteria into graded scores are never given.

3.3 The analogue series contradicts the interface analysis it is supposedly derived from, and the development plan does not match the stated indications

Table 4 defines nine constructs, of which only HSP-3 is examined computationally, and several rationales run against Supplementary Table S1. HSP-4 is described as adding Aib at three alanines “without touching any receptor-contact side chain,” yet Ala8, Ala15 and Ala18 bury 36.7, 18.3 and 38.0 Ų respectively, all above the paper's own threshold, and the same row's risk column concedes the point. HSP-8 methylates Lys16, which buries 83.7 Ų and makes three polar contacts, described as merely “close to the interface.” HSP-6 is described as acylating “a tail lysine” when the scaffold has none, requiring an additional S30K substitution that the “one substitution, or none” framing omits. Separately, protease resistance is the principal pharmaceutical claim for the macrocycle yet is never tested, modelled or connected to the accessibility data already computed, and Table 7 is a motif-derived list that could easily be cross-referenced against relative SASA. The development plan has three further gaps: there is no in vivo female efficacy model or endpoint despite equal billing for female sexual arousal disorder; there is no pharmacokinetic step corresponding to the requirement that the peptide be cleared quickly, with the plasma-stability criterion in fact pointing the other way; and immunogenicity is not mentioned anywhere, although the exendin precedent on which the translational argument rests is associated with substantial treatment-emergent anti-drug antibody rates attributed to non-human origin.

Minor Issues

1. Integrator and protocol settings. Section 2.3 reports hydrogen mass repartitioned to 1.5 amu with a 2 fs timestep, a combination that confers little benefit over standard masses, and Section 2.7 does not state the repartitioning used in production. Neither section states whether an equilibration period was discarded before the whole-trajectory averages in Section 3.8, which matters because both peptides are released from a restrained non-equilibrium pose. The choice of 350 K is also asserted without justification and without a temperature series.

2. HSP-3b is never built. The zero-substitution i,i+3 Lys21–Glu24 lactam is presented as the option that may make the lead substitution unnecessary, and is arguably the better lead once the selectivity attribution is corrected, yet it receives neither explicit construction nor a stress test. Given that the headline claim is a design costing one substitution or none, the “none” branch deserves the same treatment as the “one” branch.

3. Data and code availability. Fifteen supplementary files and four scripts are listed with no repository, accession or persistent identifier. For a study in which every claim resides in those files, a versioned archive is essential and is required by most journals in this area.

4. Figure 4 is absent. The text and caption describe the per-residue burial plot with red arcs marking candidate bridges, but no image accompanies it, unlike Figures 1, 2, 3 and 5. This is the figure carrying the central structural result.

5. Citation–claim mismatches. References 4 and 7 are cited for the presence of VIP-immunoreactive fibres in penile and clitoral tissue, but concern α-adrenergic receptors in the penis and an NO-like NANC mediator respectively. Said and Mutt (1970) is cited for VPAC1/VPAC2 class B pharmacology and for parallel NO–cGMP signalling, both of which postdate it. Reference 8, on neutral-endopeptidase inhibitors for female sexual arousal disorder, appears in the list but is cited nowhere.

6. A pull-quote contradicts the text. The standalone line preceding Figure 2 states that Lys21 “had to be engineered into VIP,” which is the opposite of what Section 3.2 and the Figure 2 caption take care to establish. It should be deleted or rewritten.

7. Table 2 and Table 4 anomalies. Helospectin I, helospectin II and helodermin are assigned identical isoelectric points and identical net charges to two decimal places despite different lengths and 11% core divergence, which suggests a calculation or copy error. Table 4's columns are transposed in several rows, and the lead compound's sequence is given only as an ellipsis.

8. PAC1 is absent from the analysis. Cascade step 1 measures cAMP at three receptors, but PAC1 appears nowhere in the modelling, the conservation scan or the selectivity comparison, despite the core template being the endogenous PAC1 agonist. Off-target PAC1 engagement is a foreseeable liability.

9. Unreferenced decision criteria. The go/no-go thresholds of a less than 15% fall in mean arterial pressure and a therapeutic index of at least 10 are numerically specific and neither is derived nor benchmarked against an existing intracavernosal comparator. The claim that systemic effects precede genital effects is likewise unsupported.

10. Selectivity-analysis reporting. Figure 7b's caption names eight residues while the text names nine; the 0.74 Å core superposition is offered as evidence that the transplant is reasonable without a threshold and measures agonist rather than pocket similarity; polar contacts of 16 against 14 are reported at a single hard cut-off; PDB 7WBJ, which would serve as a within-receptor control, is unused; and the Abstract omits Section 3.9's own disclaimer that no affinity difference should be claimed.

11. Design details requiring support. The UniProt evidence code and primary reference for the Ser32 O-glycosite underlying HSP-7 are not given. No permeation strategy is offered for the topical, intraurethral and clitoral routes proposed for a roughly 4 kDa hydrophilic peptide. The synthetic feasibility of on-resin side-chain macrolactamisation on a 38-mer, including the aspartimide risk at Asp3 that Table 7 itself flags, is asserted rather than assessed. And the additive-hypotension risk of combining a cAMP-dependent vasodilator with PDE5 inhibition is not discussed although step 3 proposes the combination.

12. Front- and back-matter defects. The manuscript lists one author but the contributions statement contains an incomplete sentence and refers to “both authors”; the by-line and correspondence give different names; the AI-assistance acknowledgment is garbled. Section 3.6 opens with a sentence fragment, a stray annotation line sits between Figure 7 and its caption, and the Discussion's selectivity paragraph contains a broken sentence with a stranded citation. The degree of overlap with the author's own preprint (reference 19) should also be characterised in a submission note.

Missing and Inadequately Integrated Literature

Cited but Inadequately Integrated

Several cited works are used as coordinate or parameter sources when their scientific content bears directly on the argument. Xu et al. (2022) is mined for the 7VQX coordinates, but its principal finding, that the VPAC2 N-terminal α-helix adopts a distinctive conformation inserting into a cleft between PACAP27 and extracellular loop 1 and that perturbing it markedly reduces cAMP accumulation, is a first-order test of whether helospectin is compatible with the threaded pose and a candidate structural basis for subtype discrimination that a contact-level analysis would miss. Piper et al. (2022) is cited solely for 8E3Z, although the same study reports the matched PACAP27-bound VPAC1 complex that would repair the confound in Section 3.9 and offers its own structural account of VPAC-family selectivity. Nguyen, Roe and Simmerling (2013) and Maier et al. (2015) are cited as the solvent model and force field without any discussion of what their validation regimes imply for a helix-retention comparison at 350 K, which is the manuscript's sole dynamical readout.

On the pharmacological side, Gourlet et al. (1997) is cited for Ro 25-1553's selectivity but its quantitative findings are not used, so Table 4's unreferenced “~100x” figure stands in place of the reported potency ratios and their species dependence. Bolin et al. (1995) is the design paper for the reference analogue and is the natural benchmark for the claim that helospectin is “already part-designed,” yet it is cited only in passing. Parker et al. (1984) is cited as historical background when it is the authority that would resolve the sequence discrepancy. Lau et al. (2015) is cited for the acylation chemistry of HSP-6 but not for its Aib-based DPP-4 strategy, which is exactly what HSP-4 proposes and what Table 7 flags at Ser2. And Laburthe and Couvineau (2002) contains the subtype-resolved distribution data on which the Discussion's entire safety argument depends, but is cited only in passing.

Truly Missing Papers

Four omissions are consequential. Tsueshita et al. (2004) reports that helospectin I and II evoke potent vasodilation in the intact hamster cheek-pouch microcirculation that is unaffected by VIP10–28, PACAP6–38 or indomethacin, and concluded the response was not transduced by VIP or PACAP receptors; this tests the manuscript's central premise directly and must be addressed. Moreno et al. (2000) is the experimental dissection assigning Ro 25-1553's VPAC2 selectivity to N-terminal acetylation, Lys12 and the C-terminal extension rather than to the 21–25 lactam, and describes a selective antagonist lacking the ring. Gourlet et al. (1998) provides the primary comparative pharmacology that the manuscript currently takes from annotation, including the rat–human divergence that bears on the proposed rodent models. And Giordanetto et al. (2013) is the closest experimental precedent for the whole design, comparing lactamisation with olefin stapling on VIP derivatives and reporting improved potency and helicity but no improvement in proteolytic stability or permeability, which speaks directly to the expected gains claimed for HSP-3 and HSP-5.

Four further works would strengthen specific sections. Duan et al. (2020) provides a second PACAP27-bound VPAC1 structure for the selectivity comparison. Dinsmore and Wyllie (2008) describes the closest existing clinical product, a VIP–phentolamine intracavernosal combination, against which the target profile should be defined. Fineman et al. (2012) quantifies the immunogenicity of the exendin precedent and is the natural source for the immunogenicity discussion the manuscript lacks. Sakamoto et al. (2021) is the source for KS-133, named as a precedent in Table 6 without a citation, and its bicyclization-derived plasma-stability data are relevant to the manuscript's own stability expectations. For the female indication, receptor-level evidence exists that would replace argument by anatomical analogy: molecular and functional work on human vaginal tissue reporting VPAC1 and VPAC2 transcripts together with VIP-mediated relaxation of precontracted smooth muscle, and rodent work reporting VIP receptor transcripts in the clitoris with VPAC2 message varying across the oestrous cycle, the latter raising a hormonal-state dependence the manuscript does not consider.

Reference list

Bolin, D. R., J. Michalewsky, M. A. Wasserman, and M. O'Donnell. 1995. Design and development of a vasoactive intestinal peptide analog as a novel therapeutic for bronchial asthma. Biopolymers 37:57-66.
Dinsmore, W. W., and M. G. Wyllie. 2008. Vasoactive intestinal polypeptide/phentolamine for intracavernosal injection in erectile dysfunction. BJU International 102:933-937.
Duan, J., D.-D. Shen, X. E. Zhou, et al. 2020. Cryo-EM structure of an activated VIP1 receptor-G protein complex revealed by a NanoBiT tethering strategy. Nature Communications 11. doi:10.1038/s41467-020-17933-8.
Endogenous vasoactive peptides and the human vagina - a molecular biology and functional study. 2010. Journal of Sexual Medicine. doi:10.1111/j.1743-6109.2010.01923.x
Fineman, M. S., K. F. Mace, M. Diamant, et al. 2012. Clinical relevance of anti-exenatide antibodies: safety, efficacy and cross-reactivity with long-term treatment. Diabetes, Obesity and Metabolism 14:546-554.
Giordanetto, F., J. D. Revell, L. Knerr, M. Hostettler, A. Paunovic, C. Priest, A. Janefeldt, and A. Gill. 2013. Stapled vasoactive intestinal peptide (VIP) derivatives improve VPAC2 agonism and glucose-dependent insulin secretion. ACS Medicinal Chemistry Letters 4:1163-1168.
Gourlet, P., P. Vertongen, A. Vandermeers, et al. 1997. The long-acting vasoactive intestinal polypeptide agonist RO 25-1553 is highly selective of the VIP2 receptor subclass. Peptides 18:403-408.
Gourlet, P., A. Vandermeers, J. Van Rampelbergh, P. De Neef, J. Cnudde, M. Waelbroeck, and P. Robberecht. 1998. Analogues of VIP, helodermin, and PACAP discriminate between rat and human VIP1 and VIP2 receptors. Annals of the New York Academy of Sciences 865:247-252.
Laburthe, M., and A. Couvineau. 2002. Molecular pharmacology and structure of VPAC receptors for VIP and PACAP. Regulatory Peptides 108:165-173.
Lau, J., P. Bloch, L. Schaffer, et al. 2015. Discovery of the once-weekly glucagon-like peptide-1 (GLP-1) analogue semaglutide. Journal of Medicinal Chemistry 58:7370-7380.
Maier, J. A., C. Martinez, K. Kasavajhala, L. Wickstrom, K. E. Hauser, and C. Simmerling. 2015. ff14SB: improving the accuracy of protein side chain and backbone parameters from ff99SB. Journal of Chemical Theory and Computation 11:3696-3713.
Moreno, D., P. Gourlet, P. De Neef, J. Cnudde, M. Waelbroeck, and P. Robberecht. 2000. Development of selective agonists and antagonists for the human vasoactive intestinal polypeptide VPAC2 receptor. Peptides 21:1543-1549.
Neuropeptide receptor transcripts are expressed in the rat clitoris and oscillate during the estrus cycle in the rat vagina. 2005. PMID 16023261.
Nguyen, H., D. R. Roe, and C. Simmerling. 2013. Improved generalized Born solvent model parameters for protein simulations. Journal of Chemical Theory and Computation 9:2020-2034.
Parker, D. S., J. P. Raufman, T. L. O'Donohue, M. Bledsoe, H. Yoshida, and J. J. Pisano. 1984. Amino acid sequences of helospectins, new members of the glucagon superfamily, found in Gila monster venom. Journal of Biological Chemistry 259:11751-11755.
Piper, S. J., G. Deganutti, J. Lu, et al. 2022. Understanding VPAC receptor family peptide binding and selectivity. Nature Communications 13:7013.
Sakamoto, K., L. Chen, T. Miyaoka, et al. 2021. Generation of KS-133 as a novel bicyclic peptide with a potent and selective VIPR2 antagonist activity that counteracts cognitive decline in a mouse model of psychiatric disorders. Frontiers in Pharmacology 12:751587.
Tsueshita, T., H. Onyuksel, V. Sethi, et al. 2004. Helospectin I and II evoke vasodilation in the intact peripheral microcirculation. Peptides 25:65-69.
Xu, Y., W. Feng, Q. Zhou, et al. 2022. A distinctive ligand recognition mechanism by the human vasoactive intestinal polypeptide receptor 2. Nature Communications 13:2272.

Conclusion and Path Forward

The question this manuscript asks is a good one, and the observation at its centre is genuinely novel: a peptide that Heloderma venom makes appears to present, without modification, several of the features that were installed synthetically in VIP to obtain a long-acting subtype-preferring agonist, and the remaining constraint can be placed on a face of the helix that the structural model assigns no role in receptor binding. The methodological instinct behind this, that a constraint should be sited by structural footprint rather than by sequence conservation, is correct and is illustrated persuasively. The authors' willingness to report their own limitations, and to present the metric that does not flatter their design, is a genuine strength and makes the work straightforward to evaluate.

What is required is that the calculations and the record catch up with the argument. In order of value, the revision should run an independent prediction of the helospectin–VPAC2 complex and repeat the burial analysis on it, so that the footprint becomes a finding rather than a protocol artefact; repeat the receptor comparison on matched PACAP27-bound templates with local repacking and a stated noise floor; run replicate dynamics including a linear [Asp25] control and either a properly parameterised isopeptide or a sham-restraint control; reconcile the sequence, the two burial columns, and Tables 3 and 4 so that the design can be checked; and integrate the primary pharmacology of helospectin and of Ro 25-1553, re-scoping the selectivity claims to what that literature supports and promoting receptor-expression profiling to the first step of the cascade. The design and development sections would then need corresponding adjustment: annotate each proposed modification with its burial value, cross-reference the protease map against computed accessibility, and add the female efficacy arm, the pharmacokinetic step and the immunogenicity assessment that the stated indications and product concept require. A revised manuscript making the more modest claim that a structurally economical macrocyclization strategy exists for a peptide whose vasodilator mechanism remains to be established, together with the experiments that would establish it, would be both more defensible and more useful than the present framing, and the field does need vasodilators that act independently of endothelial nitric oxide.