{"id":"f903ff43-2685-4d5e-946f-342c5e91ad51","arxiv_id":"2608.04834","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"The paper derives histogram-based binding-affinity estimators with explicit volume corrections, showing standard single-bin estimators miss by about 1 kcal/mol and bound-state boundaries shift transporter affinities by about -0.6 kcal/mol.","lead":"Molecular simulations can estimate binding affinities from the relative time a drug spends bound versus free. This paper derives the precise volume corrections needed for that estimate and shows that a common shortcut misreports the affinity by about 1 kcal/mol, a decisive margin in drug design.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"DgoT absolute affinity hinges on unverified geometric volume; no multi-R check of the homogeneous-unbound assumption for the smooth restraint.","rationale":"The reader identified the homogeneous-unbound-state assumption (SI S1.2.2 Eq S69) as the weakest point; my concern is a specific and more operational instance of that same assumption. The derived estimators require not only ρ̄(z)=const (the 1D plateau) but also a uniform distribution over the restraint's cross-section, and they convert geometric restraint dimensions directly into V_R. The paper's own SI footnote concedes that real harmonic restraints are only approximations to the ideal reflective wall and that flatness inside the accessible region is required. The CB7 application provides a partial test: the near-independence of ΔF0 on R in Table 2 supports the assumption, although the residual spread (0.18 kcal/mol) is larger than the formal error bars, suggesting a small systematic effect. The DgoT application provides no such test: one radius, one unbound window, and a restraint whose functional form and force constant are not given (only a reference to Ref [32]). Thus the absolute GAL–DgoT affinity is conditional on the geometric accessible volume πR²ℓ being correct. The central quantitative claim about single-bin estimators (~1 kcal/mol) is not threatened, since that discrepancy arises from the bound-state definition (single bin at the minimum undercounts a structured bound basin), and it is corroborated by the LJ dimer (Table S1, Estimate 3). But the 'exact volumetric terms' headline claim is under-supported for the biologically relevant case. A multi-R scan and a radial-density check would settle whether the concern lands. This does not change the reader's CONDITIONAL verdict; it sharpens the condition.","tokens_in":31520,"tokens_out":17556,"duration_ms":185296,"concrete_test":"Repeat the GAL–DgoT well-tempered metadynamics with two additional cylinder radii (R=0.2, 0.3 nm) and recompute ΔF0 over the same unbound window z∈[3.8,4.2] nm; accept the geometric-volume correction only if the three values agree within statistical error. Additionally, compute the ligand center-of-mass radial density within the cylinder in the plateau region (z∈[3.8,4.2] nm); if the density is not uniform, replace V_R by ∫ exp(-βV_rest(r)) d³r over the unbound window and re-evaluate Table 3.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The practical estimators reduce to K = P_R_b/(P_R_u/V_R) only under the homogeneous-unbound-state assumption (SI S1.2.2, Eq S69), and the cylindrical version further requires the ligand-COM density to be uniform over the restraint cross-section so that V_R = Σℓ is the correct accessible volume. The manuscript itself flags the soft-wall approximation (SI S1.2 footnote): a harmonic wall is only an approximation to the ideal reflective boundary, and it \"must be flat (zero force) throughout the accessible region.\" For the CB7 system this approximation is internally tested: Table 2 shows ΔF0 nearly R-independent across R=0.1,0.2,0.3 nm, though with a residual 0.18 kcal/mol spread exceeding the reported statistical errors (≈0.02–0.04 kcal/mol). For the GAL–DgoT system no such test exists: the restraint is described only as a \"smooth confining potential\" from Ref [32] with R=0.1 nm, and V_R is taken as the geometric volume πR²ℓ. If the smooth potential is not flat in the interior or has a non-negligible boundary layer, the true accessible volume ∫ exp(-βV_rest) d³r differs from πR²ℓ, biasing ΔF0. The reported DgoT value (-1.1 kcal/mol) is therefore conditional on an untested geometric assumption, and the \"exact\" volumetric term is not demonstrated for this system. The ~1 kcal/mol single-bin discrepancy is likely robust because it is driven by the bound-state definition (single bin at the minimum vs integrated bound basin), but the absolute affinity and the exactness claim for the volume term are not settled.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper derives the relation K = Pb/(Pu/V) from the canonical partition function in the dilute limit and uses it to construct practical estimators for binding free energies from histograms of a reaction coordinate. The estimators cover unrestrained and restrained simulations in spherical, circular, and cylindrical geometries, with analytical volumetric corrections. The framework is applied to a Lennard-Jones dimer, the CB7/1-adamantanol host–guest complex, and the galactonate–DgoT transporter system. The central numerical finding is that single-bin estimators deviate from integrated extended-region or hybrid estimators by about 1 kcal/mol in both molecular complexes. The paper also discusses how the bound-state definition should be matched to experimental resolution and introduces a distinction between full-access and partial-access restraining settings.","tokens_in":31814,"tokens_out":14511,"duration_ms":163749,"significance":"If the derivation is accepted, the paper offers a transparent, unified route from population histograms to absolute binding free energies, making the volumetric term explicit and the bound-state definition a controlled input. The Lennard-Jones test with three restraint radii is a clean verification that the volumetric correction cancels the restraint dependence exactly. The demonstration that single-bin estimators can be off by about 1 kcal/mol is practically important for the many methods that use such estimators. The discussion of full-access versus partial-access settings, and the resulting conditional nature of some predictions, is honest and useful. The main weaknesses are a rigor gap in the derivation of Eq. (10) and an untested geometric-volume assumption in the DgoT application; both affect load-bearing claims rather than presentation.","major_comments":[{"comment":"The derivation of Eq. (10) passes from the solvent average in Eq. (S46) to the pointwise identity in Eq. (S49) by asserting that the integrand is constant over the support of the solvent distribution. This replaces ⟨log f⟩ with log⟨f⟩ without a rigorous justification: the single-molecule partition functions z_A(S), z_B(S), and z_AB(S) are explicitly S-dependent, and the support of ρ(S) contains many solvent configurations. The final relation is standard and likely correct, but as written the central 'exact' claim rests on an unjustified step. Please either provide a thermodynamic-limit argument showing that fluctuations of the integrand vanish, or derive Eq. (10) directly from the ratio of the averaged partition functions, or qualify the exactness claim accordingly.","section":"SI S1.1, Eqs (S46)–(S51)"},{"comment":"The DgoT estimator uses the geometric accessible volume V_R = πR²ℓ with R = 0.1 nm, which is exact only if the smooth confining potential is flat inside the cylinder and the unbound density is uniform over the whole cross-section. Unlike the CB7 system, no multi-radius test is reported for DgoT, and the functional form of the smooth potential is not given. The SI footnote in S1.2 explicitly acknowledges that a harmonic soft wall is only an approximation to the ideal reflective boundary and must be flat throughout the accessible region. The reported ΔF0 ≈ −1.1 kcal/mol therefore inherits an unquantified systematic error from this assumption. I request either a direct computation of the effective volume ∫e^{−βV_rest} d³r, a second restraint radius, or an explicit limitation statement with an estimated bound on the induced error.","section":"Galactonate–DgoT; SI S1.2.2, Eq (S92)"},{"comment":"The extended-region estimates for CB7 are −17.639(4), −17.67(2), and −17.49(2) kcal/mol for R = 0.1, 0.2, and 0.3 nm. The spread of about 0.18 kcal/mol is several times the reported statistical uncertainties (0.01–0.04 kcal/mol). The text describes these values as consistent, but the residual scatter suggests a small systematic component not captured by the geometric volume correction, possibly from soft-wall penetration or an imperfect plateau. Please discuss this residual explicitly and quantify its implications for the claim that the volumetric correction is exact.","section":"Table 2"}],"minor_comments":[{"comment":"The symbol F̃(x) is introduced for −kT log ρ(x) and then not used again; consider removing it or using it consistently when discussing the dimensional ambiguity of the profile.","section":"Eq (14)"},{"comment":"The 'smooth confining potential' taken from Ref. [32] is not described in this paper; please provide its functional form or cite the specific section of Ref. [32] where it is defined, so that the geometric volume V_R = πR²ℓ can be checked against the actual potential.","section":"Computational Details, galactonate–DgoT"},{"comment":"The choice of the single-bin unbound reference z_u* should be justified as lying in a genuinely flat region of the plateau; for DgoT, z_u* = 4.0 nm lies at the upper edge of the stated unbound window [3.8, 4.2] nm, so a brief sensitivity check would strengthen the result.","section":"Hybrid estimator, Eqs (31)–(32)"},{"comment":"There is a typo: 'implememnted' should be 'implemented' in the description of the Particle Mesh Ewald scheme.","section":"SI S2.2"},{"comment":"The last row of Table 3 shows a −0.6 kcal/mol shift when the bound-state definition is changed from the primary pose to the complementary region; this is discussed clearly, but it would help to state explicitly that this shift is of the same order as the single-bin discrepancy and therefore the absolute value of the DgoT affinity remains conditional on the experimental definition of the bound state.","section":"Discussion, full-access vs partial-access"}],"recommendation":"major_revision","confidential_remarks":"The methodological advance over existing PMF-based formalisms is incremental: Eq. (10) is a standard result, and the practical estimators largely recover the structure of earlier funnel-metadynamics expressions. The paper's main contribution is the explicit comparison of estimators and the demonstration that single-bin choices introduce ~1 kcal/mol errors. The SI derivation gap (pointwise identity) and the untested geometric-volume assumption for DgoT should be fixed before publication. The CB7 residual R-dependence, while small, should also be addressed because it bears on the exactness claim. Overall the paper is within scope for the journal and likely publishable after revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a useful methods paper with one overclaim and one under-tested application. The central message—that common single-bin estimators for absolute binding affinities miss about 1 kcal/mol because of how the bound state is defined—is correct and worth taking seriously. The paper also gives a clean derivation of the volumetric correction for restrained simulations, and the hybrid estimator with explicit bin-width correction is a genuinely useful addition. But the word 'exact' is doing too much work, and the DgoT absolute affinity rests on an unverified geometric volume.\n\nWhat's new: the population-based relation K = Pb/(Pu/V) is acknowledged to be Woo-Roux's Eq (2), so that's not the contribution. The contributions are (i) the hybrid estimator that combines an integrated bound state with a single-bin unbound reference, with the Δz term made explicit; (ii) the full-access/partial-access distinction, which clarifies when a pose-specific binding constant is the right target; and (iii) a careful demonstration that single-bin estimators are off by ~1 kcal/mol in two realistic systems. The LJ dimer test and the three-radius CB7 check are well designed and support the volumetric correction.\n\nThe soft spots: the SI derivation of Eq (10) goes through a pointwise-identity step (Eqs S46–S51) that is not fully rigorous; the final relation is standard, but calling the route 'exact' overstates it. More importantly, the DgoT application uses a smooth confining potential at a single radius, and the paper's own SI footnote warns that a soft wall must be flat over the accessible region. That condition is never checked for DgoT, so the absolute ΔF0 of −1.1 kcal/mol is conditional on the geometric volume πR²ℓ being the true accessible volume. This does not affect the main single-bin discrepancy, which is driven by the bound-state definition, but it means the 'exact volumetric terms' are not demonstrated for that system. Also, no code or data is deposited, which is a real omission for a methods paper.\n\nFor whom: practitioners who compute absolute binding affinities from histograms or PMFs. They will get a clear warning about single-bin estimators and a useful reference for how to set up the volumetric correction. It deserves a serious referee; I would ask for a revision that adds a multi-radius test (or a direct check of restraint flatness) for DgoT, tones down 'exact', and deposits the inputs and scripts.","headline":"Solid methods paper with a useful warning about single-bin estimators, but the 'exact' framing and the DgoT volume need qualifying.","tokens_in":32380,"tokens_out":3867,"would_cite":true,"duration_ms":40915,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper derives an exact relation between binding constants and population weights, Eq.","keywords":["binding free energy","binding constant","statistical weight","volumetric correction","reaction coordinate histogram","well-tempered metadynamics","cucurbit[7]uril","ligand-protein binding"],"falsifier":"Take any restrained binding simulation with a known experimental affinity and check whether the unbound density becomes exactly flat over the full interval used for the volume correction, and whether moving the unbound window within that plateau changes the final free energy by more than the statistical uncertainty; if it does, the homogeneous-unbound assumption is the cause. Alternatively, re-bin a fixed trajectory at two different bin widths and verify that the hybrid and single-bin estimators shift by exactly the predicted $-k_BT\\log(\\Delta z)$ term, while the extended-region estimator does not.","tokens_in":31285,"feed_emoji":"⚛️","tokens_out":6476,"duration_ms":68315,"temperature":0.7,"pith_summary":"The paper's central claim is that the equilibrium binding constant in the dilute limit is exactly $$K = \\frac{P_b}{P_u/V},$$ where $P_b$ and $P_u$ are the statistical weights of the bound and unbound states defined from the canonical density, and $V$ is the accessible volume. The authors derive this from the partition function in the explicit-solvent setting, which brings the unbound-state volumetric term out of the code and shows exactly how a volume restraint propagates into the affinity. From this single relation they build practical estimators that work off histograms of any reaction coordinate, in unrestrained or restrained simulations, with closed-form corrections for spherical, circular, and cylindrical geometries. Applied to the CB7/1-adamantanol host–guest complex and the galactonate–DgoT transporter, the theoretically grounded estimators agree with each other while the commonly used single-bin estimator deviates by about 1 kcal/mol. A sympathetic reader would take the paper to establish that the definition of the bound state and the unbound volume are controlled inputs, not hidden assumptions.","feed_headline":"Exact binding constants from one population ratio","feed_subtitle":"Single-bin histogram estimators miss by ~1 kcal/mol; the correct volume term is now explicit.","key_machinery":"The load-bearing object is the statistical weight $P_s = \\int \\rho(q,p)I_s(q,p)\\,dq\\,dp$, the integral of the canonical phase-space density over the region defining a thermodynamic state. The identity $K = P_b/(P_u/V)$ carries the argument: it turns a binding constant into a ratio of populations, so every volumetric effect reduces to the accessible volume of the unbound state, and an imposed restraint enters simply as $V \\to V_R$. The practical estimators then run on histograms of a reaction coordinate $z$ or $r$, kernel density estimates, and pair correlation functions $g_3(r)$ or $g_2(r)$, which supply the convergence diagnostics (a flat plateau in $\\bar\\rho(z)$ or $g_3(r)\\to 1$) that certify the homogeneous-unbound-state assumption.","core_discovery":"The central discovery is that Eq. (10), $K = P_b/(P_u/V)$, is exact under the stated dilute-limit assumptions, and that all practical binding-affinity estimators follow from it. The proof starts from the canonical partition function for $N_A$ ligands, $N_B$ receptors, and solvent, factorizes it in the dilute limit, and identifies the statistically dominant term; the equilibrium condition then yields the ratio of statistical weights. The volume dependence lives entirely in the unbound state, so the standard correction $-k_BT\\log(VC_0)$ is not an appended term but a direct consequence of the normalization of $P_u$. For restrained simulations, the same relation becomes $K = P_b^R/(P_u^R/V_R)$ under the physically mild assumption that the unbound density is homogeneous over the restrained region, and the paper derives explicit $V_R$ formulas for spherical, circular, and cylindrical restraints. The applications show that integrated bound-state estimators (extended-region and hybrid) give mutually consistent free energies across restraint sizes, while the single-bin estimator consistently departs by about 1 kcal/mol, and that in a structured bound state the placement of the state boundary can shift the affinity by about $-0.6$ kcal/mol.","pith_inferences":["The construction is not tied to two-body complexes: the same population-ratio logic should yield exact volumetric terms for dimerization equilibria or multi-site binding, where the unbound state has a geometric volume (or area) of its own.","The size of the single-bin error is tied to how much bound-state weight a single histogram bin captures; for flexible ligands with multiple poses the deviation could plausibly exceed the 1 kcal/mol seen here.","A practical by-product is a consistency test for existing PMF-based pipelines: any implementation whose unbound reference is not the true accessible unbound volume should show a residual dependence of $\\Delta F_0$ on the restraint radius, which this formalism can detect."],"forward_implications":["Absolute binding affinities can be computed from existing PMF or histogram data by reweighting the bound and unbound populations, without additional simulations.","Single-bin estimators should be avoided whenever the bound state has internal structure; in the two molecular systems tested they are off by approximately 1 kcal/mol.","The volume in the standard-state correction is the unbound-state accessible volume, so restraint geometries must be designed so that $V_R$ (or $\\Sigma\\ell$) is well defined and the restraint is inactive in the bound state.","Comparing computed affinities to bulk equilibrium experiments requires an inclusive bound state; site-resolved experiments may validly target a single sub-basin, in which case a restricted definition is the correct target."],"supporting_citations":[{"why":"Supplies the statistical-thermodynamic basis for binding constants that the derivation reformulates in terms of population weights.","marker":"[11]"},{"why":"Gives the prior Eq. (2) equivalent to Eq. (10); the paper rederives it from the partition function rather than from the standard single-receptor expression.","marker":"[19]"},{"why":"A PMF-based funnel metadynamics estimator whose structure the present derivation recovers and whose bound-state definition it makes explicit.","marker":"[22]"},{"why":"A volume-based metadynamics estimator the paper's framework places in a unified derivation.","marker":"[23]"},{"why":"Provides the CB7/1-adamantanol force-field setup and prior free-energy results used as comparison.","marker":"[29]"},{"why":"Source of the GAL–DgoT simulation trajectory data on which the protein application is based.","marker":"[32]"},{"why":"Supplies the time-independent reweighting used to turn well-tempered metadynamics samples into unbiased histograms.","marker":"[44]"},{"why":"The well-tempered metadynamics method used to generate the enhanced-sampling trajectories.","marker":"[49]"}],"fun_headline_variants":["Exact binding constants from one population ratio","Why single-bin estimators get affinity wrong by 1 kcal/mol","Volumetric terms made explicit for absolute affinities","A principled route to exact binding constants in simulations"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The derivation's load-bearing premise is that the unbound state is homogeneous, so the probability density is constant over the entire restricted region; if residual correlations remain in that region, replacing the unbound integral with a constant density times the volume is inexact.","fun_headline_variants_meta":{"raw":{"variants":["Exact binding constants from one population ratio","Why single-bin estimators get affinity wrong by 1 kcal/mol","Volumetric terms made explicit for absolute affinities","A principled route to exact binding constants in simulations"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000302,"raw_usage":{"total_tokens":1782,"prompt_tokens":1033,"completion_tokens":749,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":649,"completion_tokens_details":{"reasoning_tokens":686}},"tokens_in":649,"tokens_out":749,"duration_ms":8349,"temperature":1.0,"reasoning_tokens":686,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T15:25:01.755043+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take any restrained binding simulation with a known experimental affinity and check whether the unbound density becomes exactly flat over the full interval used for the volume correction, and whether moving the unbound window within that plateau changes the final free energy by more than the statistical uncertainty; if it does, the homogeneous-unbound assumption is the cause. Alternatively, re-bin a fixed trajectory at two different bin widths and verify that the hybrid and single-bin estimators shift by exactly the predicted $-k_BT\\log(\\Delta z)$ term, while the extended-region estimator does not.","supporting_citations":[{"cited_title":"The Statistical- Thermodynamic Basis for Computation of Binding Affinities : A Critical Review.Biophysical Journal, 72(3):1047–1069, 1997","cited_arxiv_id":null,"evidence_quote":"Supplies the statistical-thermodynamic basis for binding constants that the derivation reformulates in terms of population weights."},{"cited_title":"Funnel metadynamics as ac- curate binding free-energy method.Proceedings of the National Academy of Sciences, 110(16): 6358–6363, 2013","cited_arxiv_id":null,"evidence_quote":"A PMF-based funnel metadynamics estimator whose structure the present derivation recovers and whose bound-state definition it makes explicit."},{"cited_title":"Exhaustive search of ligand binding pathways via volume-based metadynamics.The Journal of Physical Chemistry Letters, 10(12): 3495–3499, 2019","cited_arxiv_id":null,"evidence_quote":"A volume-based metadynamics estimator the paper's framework places in a unified derivation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the CB7/1-adamantanol force-field setup and prior free-energy results used as comparison."},{"cited_title":"Protonation-dependent substrate release in a bacterial homolog of vesicular glutamate.Biophysical Journal, 125(7):1565–1569, apr 2026","cited_arxiv_id":null,"evidence_quote":"Source of the GAL–DgoT simulation trajectory data on which the protein application is based."},{"cited_title":"Well-Tempered Metadynamics: A Smoothly Converging and Tunable Free-Energy Method.Physical Review Letters, 100(2):020603,","cited_arxiv_id":null,"evidence_quote":"The well-tempered metadynamics method used to generate the enhanced-sampling trajectories."}],"review_version":1}