{"id":"1a439936-baeb-4ea7-8829-d190295dede7","arxiv_id":"1908.05693","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"The P23T mutation does not produce a consistent increase in γD-crystallin surface hydrophobicity, and hydrophobic-effect models require finely tuned parameters to invert solubility, so microscopic support for the hydrophobic mechanism is weak.","lead":"This paper tests whether water-hating surface patches explain why a mutant of γD-crystallin dissolves less when heated. It finds that the hydrophobicity evidence is weak and that inverted solubility requires a delicate balance of patch strengths and temperature effects.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'weak support' conclusion hinges on MLG degeneracies and unhalved patch energies; with Shiryayev degeneracies or εtot=30, n_w* drops below the estimated available water, so the key threshold is not robust.","rationale":"The reader's CONDITIONAL verdict is appropriate and should be retained. The surface-hydrophobicity comparison across five scales is a genuine negative result, and the temperature-deactivated patch model sustains inversion under ±10% patch perturbations; both pieces support the paper's caution. The load-bearing weakness is in the quantitative MLG threshold: the manuscript itself calls the degeneracies 'fairly arbitrary' (Sec. 4.1), yet n_w* swings from ~71 to ~18 across the two published choices, straddling the empirical 43–48 water estimate. Because the abstract states 'microscopic evidence to support it ... is weak' rather than 'evidence is inconclusive,' this parameter sensitivity matters for the central claim. The additional εtot=30 calibration (n_w*≈35) reinforces the concern. The text/Eq. (23) discrepancy in the inversion condition is a secondary, but easily checked, inconsistency. None of this changes the verdict: the paper is already conditional and carefully hedged, and the missing data archive/SI would need to be resolved before full acceptance. I therefore leave the CONDITIONAL verdict unchanged.","tokens_in":21152,"tokens_out":9220,"duration_ms":86007,"concrete_test":"Recompute n_w* from Eq. (23) and Fig. 5 using the Shiryayev et al. degeneracies with both εtot=60 and εtot=30, and also with the Silverstein et al. degeneracies at εtot=30, keeping all other MLG energies and patch geometry fixed. If any of these physically motivated calibrations yields n_w*≤43 (the lower end of the estimated hydrophobic-residue water count in Sec. 4.1), the abstract's 'microscopic evidence ... is weak' statement fails for that calibration, and the manuscript must either add an independent constraint selecting one degeneracy set or rephrase the conclusion as parameter-set-dependent.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central inference that hydrophobicity is only weakly supported in γD-crystallin rests on the number of water molecules n_w* required to invert solubility, Eq. (23). This threshold is not pinned down. Sec. 4.1 explicitly calls the MLG degeneracies 'fairly arbitrary': with Silverstein et al. parameters n_w*≈71, but with the Shiryayev et al. multiplicities n_w*≈18. The manuscript's own empirical estimate of water molecules solvating the hydrophobic residues of Patch 4 is 43–48 (an upper bound), so the alternative degeneracy set places the system comfortably above threshold. The same threshold drops to n_w*≈35 if the total patch energy is halved to εtot=30 to match experimental solubility, a correction the paper treats as reasonable because εtot=60 gives solubilities orders of magnitude too low. Since n_w*<43 in that calibration, the hydrophobic scenario is quantitatively sufficient under a plausible, self-consistent parameter choice. Thus 'weakly supported' is a choice among arbitrary parameter sets rather than a robust conclusion. Separately, Sec. 4.1 states the inversion condition as Γ>εtot/(2nw)−Δε(1)/nw, while Eq. (23) has Δε(1)/2; this internal inconsistency should be resolved before relying on the reported n_w* values.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper investigates whether inverted solubility in human γD-crystallin P23T mutants is caused by increased surface hydrophobicity. It first measures SASA-weighted hydrophobicity with five amino-acid scales in six crystal structures and finds no consistent, statistically significant increase associated with P23T. It then studies a patchy-particle model of the DBI and DBN crystal forms in which Patch 4, containing residue 23, has a temperature-dependent energy given by one of three models: the MLG two-state water model, the Wentzel–Gunton linear model, and a generic temperature-deactivation model. The authors derive an explicit inversion condition, Eq. (23), identify the number of waters n_w needed to invert solubility, compare it to structural estimates of available hydrophobic waters (43–48), and map solubility lines and critical temperatures. They conclude that hydrophobic-effect-driven inversion is possible only in a narrow parameter window and that microscopic support in γD-crystallin is weak, while the generic deactivation model robustly captures the phenomenology and leaves the microscopic cause open.","tokens_in":21479,"tokens_out":11556,"duration_ms":111351,"significance":"If accepted, the paper offers a useful cautionary result: surface-hydrophobicity scales do not support the earlier dye-binding inference, and the hydrophobic scenario requires a fine balance of parameters, helping explain why inverted solubility is rare. The paper is transparent: the analytical derivation of the inversion condition is explicit, the limitations of the ground-state crystal free-energy approximation are stated, and the authors flag the arbitrariness of MLG degeneracies. It also provides specific falsifiable predictions, including the ordering of hypothetical R36S+P23S/P23V double mutants and the absence of closed-loop binodals unless Patch 4 is more than doubled in strength. The main limitation is that the key threshold n_w* is sensitive to parameter choices that the paper itself acknowledges as reasonable, so the headline conclusion is more parameter-dependent than the abstract suggests.","major_comments":[{"comment":"The central conclusion that hydrophobic support is weak depends on a threshold that the paper's own parameter choices straddle. With the Silverstein et al. degeneracies and εtot=60, n_w*≈71 exceeds the estimated 43–48 available hydrophobic waters; with the Shiryayev et al. degeneracies, n_w*≈18; and with the halved patch-energy calibration εtot=30 that Sec. 4.1 calls reasonable, n_w*≈35. Because the halved calibration was introduced to bring computed solubilities into line with experiment, the hydrophobic scenario is quantitatively sufficient under at least one plausible, self-consistent parameter set. The manuscript should either justify a preferred parameter set or soften the abstract's 'weakly supported' claim to 'inconclusive'.","section":"Sec. 4.1, Eq. (23), Figs. 5 and 6"},{"comment":"The inversion condition is stated inconsistently. Eq. (23) reads Γ(β)=εtot/(2nw)−Δε(1)/2, while the text immediately below Eq. (23) states Γ(β)>εtot/(2nw)−Δε(1)/nw. Since the reported n_w* values are derived from this condition, this discrepancy needs to be resolved before the threshold analysis is reliable.","section":"Sec. 4.1 and Eq. (23)"},{"comment":"The surface-hydrophobicity comparison draws on crystal structures solved in different lattices, and the Patch 4 SASA is evaluated within the crystal context of each structure. The paper addresses coordinate uncertainty but does not discuss whether the different lattice environments of 1HK0, 4JGF, 2G98, 1H4A, 6ETA and 6ETC bias the comparison of a given surface patch. A brief discussion of this possible bias, or a restriction of the claim to the structures themselves, would strengthen the first pillar of the conclusion.","section":"Sec. 2, Fig. 1"}],"minor_comments":[{"comment":"The sentence 'Proteins to the left of the black vertical line exhibit normal solubility, and those to the left exhibit inverted solubility' should read '...and those to the right...'.","section":"Fig. 1 caption"},{"comment":"In the affiliations, 'I reland' should be 'Ireland'.","section":"Author affiliation"},{"comment":"The notation −ε′_4 = ... on the left-hand side is confusing; please define ε′_4 explicitly rather than writing the negative of the patch energy.","section":"Eq. (12)"},{"comment":"The text contains the LaTeX artifact 'n_w* /greaterorsimilar71'; it should be typeset as n_w* ≳ 71.","section":"Sec. 4.1"},{"comment":"Reference 33 is incomplete ('Schrödinger, LLC'), and the data/SI DOI placeholders 'https://doi.org/10.7924/XXXXXXX' and 'DOI: 10.1021/XXXX' need to be completed before publication.","section":"References and data availability"},{"comment":"The integral limits are written awkwardly as '4.5 Å ∫ 3 Å'; please typeset this as an integral from 3 Å to 4.5 Å.","section":"Eq. (27)"}],"recommendation":"major_revision","confidential_remarks":"The paper is honest and clearly written; the main issue is that the headline conclusion is more parameter-dependent than the abstract conveys. A major revision with a sensitivity analysis and careful rewording of the conclusions should be sufficient to address this."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this paper is a careful, honest computational test of the hydrophobicity scenario for inverted solubility in γD-crystallin, and it deserves a serious referee. The new content is not the patchy model itself but the multi-scale hydrophobicity screen and the threshold analysis (n_w*) for the MLG and Wentzel–Gunton models. The screen shows no consistent hydrophobicity increase across five scales for P23T structures, which weakens the naive story. The model work shows that hydrophobicity-induced inversion requires a fine balance of patch strengths, water count, and temperature dependence. That is a real, non-obvious result.\n\nThe paper is self-aware. It explicitly flags the MLG degeneracies as fairly arbitrary, notes that the deactivated-patch model builds inversion in by construction, and shows where its crystal free energy approximation breaks down. Those are the right things to do. The derivations are transparent, Eq. (23) is simple enough to check, and the robustness tests on patch energies and τ are useful.\n\nThe soft spot is the one you flagged. The conclusion that hydrophobicity is 'weakly supported' rests on n_w* ≈ 71 from the Silverstein degeneracies. With Shiryayev degeneracies n_w* ≈ 18, and with halved patch energies n_w* ≈ 35. Both are below the estimated 43–48 water molecules. So the quantitative threshold flips under plausible parameter choices, and the paper's own abstract makes a stronger statement than the parameter dependence really supports. They do say 'may be possible' and 'weak', but the logic should be presented as 'the available water count is near or above the threshold for some parameter sets, below for others' rather than a clean verdict. Also, Sec. 4.1 has a typo: the inequality writes Δε(1)/n_w where Eq. (23) has Δε(1)/2. Eq. (23) is the correct one from the derivation, so this is likely a typesetting slip, but it matters for anyone reading the text literally. The placeholder data DOI and the reliance on the SI of Ref. 30 are also problems for a computational paper; the archive should be live.\n\nNone of this sinks the paper. The qualitative claim—that the hydrophobic scenario is not obviously supported and that the temperature-deactivated generic model remains viable—holds up. The paper just needs to be more careful about how parameter-dependent the quantitative statement is.\n\nRecommendation: accept with minor revisions. It deserves referee time. I'd bring it to a reading group if anyone works on patchy models or protein phase behavior.","headline":"A careful, honest computational test of the hydrophobicity story for inverted solubility, but the headline conclusion is more parameter-dependent than the abstract suggests.","tokens_in":21995,"tokens_out":2910,"would_cite":true,"duration_ms":27495,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Inverted solubility in γD-crystallin is not the simple hydrophobic effect it was thought to be.","keywords":["inverted solubility","gammaD-crystallin","P23T mutation","hydrophobic effect","patchy particle model","solubility phase diagram","water solvation","temperature-dependent patch interactions"],"falsifier":"Measure the actual number of water molecules in the first solvation shell around the residue-23 patch (for example by neutron diffraction or by simulations with a more sophisticated water model) and compare it with the model's threshold: if the measured count is below $n_w^* \\approx 71$ (or $\\approx 35$ for halved patch energies), the hydrophobic-effect scenario is falsified for this protein; if it meets or exceeds the threshold, the scenario remains viable.","tokens_in":2019,"feed_emoji":"💧","tokens_out":5189,"duration_ms":108415,"temperature":0.7,"pith_summary":"The paper tests a proposed microscopic explanation for inverted solubility in human γD-crystallin, where the P23T mutation makes the crystal melt upon cooling rather than upon heating. The proposed mechanism is that the mutation increases surface hydrophobicity. Measuring five different hydrophobicity scales on available crystal structures, the authors find no consistent or statistically robust increase in hydrophobicity for structures carrying the mutation. They then ask whether a hydrophobic mechanism could still work thermodynamically by building a schematic patchy-particle model with three temperature-dependent patch-energy descriptions: two explicit hydrophobic-effect models and one generic patch-deactivation model. The paper concludes that solubility inversion from the hydrophobic effect is possible only in a narrow parameter window, that microscopic evidence for it in γD-crystallin is weak, and that a generic temperature-deactivated patch remains a viable description.","feed_headline":"Hydrophobicity alone can't explain crystallin's inverted solubility","feed_subtitle":"P23T mutants show no hydrophobicity gain; patchy models need a fine balance of patch strength and water count.","key_machinery":"The load-bearing object is a schematic patchy-particle model in which each protein is a hard sphere with a set of adhesive patches whose positions, ranges, and energies are taken from the DBI (inverted-solubility) and DBN (normal-solubility) crystal contacts; Patch 4, which contains residue 23, carries a temperature-dependent energy. Three forms for that temperature dependence are compared: the MLG four-state water model (water in ordered or disordered states in the shell or bulk, with energy and degeneracy parameters from the model), the Wentzel–Gunton linear model, and a tanh-based temperature-(de)activation model. Solubility lines come from equating the fluid chemical potential, obtained from the second virial coefficient, with the crystal chemical potential, obtained from thermodynamic integration from the crystal reference state. The argument's crux is the inversion condition of Eq. (23): the chemical-potential slope is positive, giving inverted solubility, when $\\Gamma(\\beta) > \\varepsilon_{\\rm tot}/(2 n_w) - \\Delta\\varepsilon(1)/2$, where $n_w$ is the number of water molecules solvating Patch 4 and $\\Gamma$ collects the temperature derivatives of the water-state entropy and energy. This identity determines the threshold water count and makes the conclusion hinge on $n_w$ and on the water-model degeneracies.","core_discovery":"On the authors' own terms, the central claim is that the hydrophobic-effect scenario for inverted solubility in γD-crystallin is not supported by the microscopic evidence available. Surface hydrophobicity computed from crystal structures with five different scales does not discernibly increase when residue 23 is mutated to threonine, serine, or valine. In the schematic patchy model, the two hydrophobicity-based potentials produce solubility inversion only when the number of water molecules $n_w$ solvating the residue-23 patch exceeds a threshold set by Eq. (23), $\\Gamma(\\beta) > \\varepsilon_{\\rm tot}/(2 n_w) - \\Delta\\varepsilon(1)/2$. Using the paper's chosen water-model degeneracies this threshold is $n_w^* \\approx 71$, above the estimated available count of roughly 43–48; halving the patch energies to match experimental solubility lowers it to $n_w^* \\approx 35$, just below that count. The paper therefore concludes that inverted solubility due to hydrophobicity may be possible but requires a fine balance between patch strength and the temperature-dependent contribution, which may explain why the phenomenon is rare. In the generic temperature-deactivated patch model, the inverted-solubility regime is robust to parameter perturbations, and the temperature-dependent interaction has a negligible effect on the liquid-liquid critical point.","pith_inferences":["Because the threshold $n_w^*$ depends on the water-model degeneracies, the paper's 'weak support' verdict is conditional: with the alternative degeneracy set considered in Sec. 4.1, inversion would need only about 18 water molecules, below the estimated 43–48 available, so a hydrophobicity explanation would then be viable.","A direct experimental measure of water structure around residue 23—for example neutron diffraction or hydrogen-deuterium exchange—would be a sharper test than hydrophobicity scales, because $n_w$ and its temperature dependence are the physical quantities that decide between the three models.","The same fine-balance argument may generalize to other proteins with inverted solubility: the phenomenon should be more likely when a single strong crystal-contact patch is solvated by a large, contiguous hydrophobic area, and less likely when hydrophobic residues are scattered.","A testable extension would be to tune $n_w$ experimentally with cosolutes or by mutating nearby residues to alter local water exposure; the model predicts that crossing the $n_w^*$ threshold at fixed patch strength should flip a normal-solubility mutant into an inverted one."],"forward_implications":["If the paper is right, surface hydrophobicity scales provide no consistent microscopic signature for the solubility-inverting P23T mutation, so the earlier dye-binding inference of increased hydrophobicity does not survive scrutiny.","Hydrophobic-effect models imply that inverted solubility should be rare, because it appears only when the patch strength, the number of solvating water molecules, and the temperature dependence are finely balanced.","The temperature-deactivated patch model reproduces the DBI inverted-solubility regime and remains inverted under 5–10% perturbations of patch energies, so it stays viable as a generic explanation without specifying the microscopic cause.","The temperature-dependent Patch 4 interaction leaves the liquid-liquid binodal essentially unchanged ($T_c \\approx 1.85$), matching experiments; a closed-loop binodal appears only for Patch 4 energies above $\\varepsilon_4 > 36$, far outside estimated error.","Because increasing Patch 4 strength lowers DBI solubility, the model predicts that R36S+P23S and R36S+P23V double mutants, if they crystallize with similar contacts, would show inverted solubility ordered by the strength of their residue-23 patch."],"supporting_citations":[{"why":"Supplies the DBI/DBN patchy model parameters and the original temperature-deactivated patch model that this paper extends and tests against hydrophobic models.","marker":"Ref. 30"},{"why":"Reports the dye-binding evidence that P23T mutants increase surface hydrophobicity; the paper's structure-based measurements directly test this claim.","marker":"Ref. 20"},{"why":"Provides the MLG-based phase diagram for globular proteins with isotropic hydrophobic interactions and an inverted-solubility regime.","marker":"Ref. 23"},{"why":"Introduces the Wentzel–Gunton linear temperature-dependent patch energy used here as one of the hydrophobic models.","marker":"Ref. 24"},{"why":"Formulates the hydrophobic interaction model for upper and lower critical solution temperatures that underlies the MLG water-state description.","marker":"Ref. 25"},{"why":"Develops the two-state hydrophobic hydration model with compensating enthalpy and entropy changes that the MLG model refines.","marker":"Ref. 26"},{"why":"Supplies the Mercedes-Benz water-model energy and degeneracy parameters used for the MLG calculations, including those that set $n_w^* \\approx 71$.","marker":"Ref. 50"},{"why":"Provides the water radial distribution function used to estimate the physically available number of water molecules around Patch 4.","marker":"Ref. 48"},{"why":"Provides experimental phase diagrams and binodal data for γD-crystallin mutants, used to compare critical properties and solubility ordering.","marker":"Ref. 31"}],"fun_headline_variants":["Hydrophobicity can't explain crystallin's inverted solubility","P23T crystallin: no hydrophobicity gain, so why inverted solubility?","Inverted solubility needs fine balance, not just hydrophobicity","Crystallin's odd melting: hydrophobicity not the culprit","Rare inverted solubility demands precise patch strength"],"cache_read_input_tokens":24064,"weakest_assumption_plain":"The conclusion that hydrophobic support is weak depends on which set of water-state degeneracies is used in the model; with one alternative set, the needed number of water molecules drops below the estimated available count, which would flip the conclusion.","fun_headline_variants_meta":{"raw":{"variants":["Hydrophobicity can't explain crystallin's inverted solubility","P23T crystallin: no hydrophobicity gain, so why inverted solubility?","Inverted solubility needs fine balance, not just hydrophobicity","Crystallin's odd melting: hydrophobicity not the culprit","Rare inverted solubility demands precise patch strength"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000442,"raw_usage":{"total_tokens":2315,"prompt_tokens":1097,"completion_tokens":1218,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":713,"completion_tokens_details":{"reasoning_tokens":1130}},"tokens_in":713,"tokens_out":1218,"duration_ms":7930,"temperature":1.0,"reasoning_tokens":1130,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:06:58.889709+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the actual number of water molecules in the first solvation shell around the residue-23 patch (for example by neutron diffraction or by simulations with a more sophisticated water model) and compare it with the model's threshold: if the measured count is below $n_w^* \\approx 71$ (or $\\approx 35$ for halved patch energies), the hydrophobic-effect scenario is falsified for this protein; if it meets or exceeds the threshold, the scenario remains viable.","supporting_citations":[],"review_version":1}