{"id":"c616499a-afbb-4377-a7bd-f63aa8243009","arxiv_id":"1908.02181","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Many-body van der Waals interactions destabilize isolated protein native states but stabilize them in explicit water through collective protein-water electron fluctuations.","lead":"This computational study simulates the quantum-mechanical dispersion (van der Waals) forces between three small proteins and their surrounding water. It finds that many-body, collective electron fluctuations in water stabilize folded protein states and can act over surprisingly long distances, suggesting current force fields may misdescribe protein solvation.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Gas-phase leg of the central claim is computed on water-stripped solvated snapshots, not gas-phase ensembles; the reported 4.5-6 kcal/mol many-body destabilization could shift if gas-phase relaxation changes native/unfolded structures.","rationale":"The reader's conditional verdict already captures this concern as the weakest assumption, and my analysis agrees that it is the most load-bearing issue. The paper's central claim is not merely that MBD differs from pairwise models on MD snapshots; it is a physical claim about gas-phase stability and its compensation by protein-water many-body interactions. That claim requires an appropriate gas-phase ensemble, or at least evidence that stripping water does not change the relevant energy differences. The proposed test would settle this directly. I do not think the concern warrants rejection or unverified status: the MBD formalism is established, the comparison across three proteins and multiple pairwise baselines is internally consistent, and no parameters are fitted to the folding data. The qualitative direction of the many-body correction is plausible even if the magnitude may shift. The reader's conditional verdict, pending this structural validation, remains appropriate.","tokens_in":15257,"tokens_out":8377,"duration_ms":105082,"concrete_test":"Select 10 native and 10 unfolded snapshots of Fip35-WW; remove water and run short gas-phase MD (e.g., 1 ns Langevin dynamics at 300 K) or energy minimization with the same force field, then recompute DFTB+MBD and vdW(TS) energies on the relaxed gas-phase structures. If the native-minus-unfolded MBD-vs-pairwise difference shifts by more than about 2 kcal/mol, the reported gas-phase destabilization is not robust to the structural artifact.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing weakness is the gas-phase reference used in Eq. 1 and in the first paragraph of Results. The isolated-protein energies E_vdW[p] are obtained by removing water from snapshots of explicit-solvent MD trajectories, so the 'gas phase' conformations are not equilibrium structures of the isolated protein. The central claim that many-body dispersion 'significantly decreases the relative stability of the native state in the absence of water' therefore rests on a single-point energy difference at solvated geometries. In vacuo, the protein would relax: side chains repack, backbone H-bonds and salt bridges reform, and the hydrophobic core can expand or contract; each of these changes alters the distribution of close contacts that determines the MBD-minus-pairwise correction. The size of the correction (4.5-6 kcal/mol for Fip35-WW, 4-5 kcal/mol for cln025/HP35) is comparable to the conformational relaxation energies one would expect, so it is not obviously small. The authors' own Discussion limits the work to an energetic 'first step' and calls for free-energy studies, but it does not flag the structural mismatch between the solvated geometries and a genuine gas-phase ensemble. If the true gas-phase native basin has a different packing motif, the direction and magnitude of the many-body correction could change, which would also alter the claimed compensation by protein-water many-body interactions.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript uses density-functional tight-binding combined with the many-body dispersion (MBD) formalism and with pairwise vdW models (vdW(TS), D2, D3) to evaluate van der Waals energies along pre-existing explicit-water MD folding trajectories of three small proteins (Fip35-WW, cln025, HP35-NleNle). The authors report that beyond-pairwise dispersion reduces the relative stability of the native state of the isolated protein by roughly 4–6 kcal/mol, while the many-body character of protein–water interactions increases the relative stability of native conformations in solution, with low-frequency collective electronic eigenmodes contributing substantially to solvation and correlation ranging to ~25 Å from the protein surface. The paper interprets these modes as plasmon-like solute–solvent interactions and argues that pairwise force fields miss a compensating balance between intra-protein and protein–water vdW effects.","tokens_in":15484,"tokens_out":6585,"duration_ms":70338,"significance":"The result, if reliable, is significant: it identifies a concrete, computationally testable failure mode of pairwise-additive dispersion in biomolecular modeling and suggests a physical mechanism (collective electronic fluctuations) that could be probed by THz spectroscopy. The study has notable strengths: the qualitative trend is reproduced for three proteins with different secondary structures and across three pairwise baselines; the MBD parameters and DFTB parameter sets are inherited from previous benchmarks rather than fitted to the proteins studied; and the mode-resolved decomposition (Eqs. 6–7) gives a mechanistic picture rather than a purely energetic correction. The main risk is that the central gas-phase comparison is performed at water-stripped solvated geometries, so the headline numbers may depend on conformational relaxation that the paper does not consider.","major_comments":[{"comment":"The isolated-protein energies E_vdW[p] are computed by 'artificially removing the surrounding solvent' from explicit-water MD snapshots, so the gas-phase leg of the central claim is a single-point energy difference at solvated geometries rather than an energy difference between equilibrium gas-phase conformers. The reported 4–6 kcal/mol many-body destabilization is comparable in size to expected conformational relaxation energies, so it could be altered or inverted by gas-phase relaxation; please re-optimize or shortly propagate the stripped structures in the gas phase, or otherwise demonstrate that the solvated geometries are representative of the gas-phase ensemble.","section":"Results, first paragraph; Eq. (1)"},{"comment":"For cln025 the SI states that 'we still find a slight destabilization of native states via many-body dispersion effects (2 kcal/mol)' in solvation, while the main text and Fig. 4 present the many-body protein–water contribution as consistently stabilizing native states across all systems. Because the 'independent of secondary structure' claim depends on this uniformity, the sign and decomposition of the cln025 net effect must be reconciled and stated explicitly.","section":"Supplementary §3.2 vs. Fig. 4"},{"comment":"The claim that electronic correlation 'is still relevant for the protein-water interaction' up to 25 Å from the interface is not quantifiable as stated. Long-range dipole-coupled models have nonzero contributions at all distances; without an explicit threshold, convergence criterion, or a defined fraction of the integrated G_int, the 25 Å value is not a falsifiable prediction. Please define 'relevant' quantitatively and report the integrated fraction as a function of cutoff.","section":"Fig. 3b and 'Plasmon-like character' section"},{"comment":"No statistical uncertainties or block-averaging estimates are reported for the 2–7 kcal/mol differences between MBD and pairwise models, and the trajectories are single runs per protein. As these differences are the basis for the qualitative conclusions, please provide at least an estimate of the noise from conformational subsampling or replicate trajectories.","section":"Results and Supplementary Material"}],"minor_comments":[{"comment":"The statement that vdW interactions 'can account for up to 30% of the total solvation energy' appears without a direct derivation in the Results; either locate the number in the figures or SI, or qualify it as an estimate.","section":"Introduction"},{"comment":"Define explicitly that 'gas phase' means the protein geometry extracted from the solvated trajectory, and use a notation such as E_vdW[p(solv)] to avoid implying a gas-phase ensemble.","section":"Eq. (1)"},{"comment":"The 'average over-stabilization' values differ between the main text ('6 kcal/mol on average') and the SI (6 kcal/mol for vdW(TS), 4 kcal/mol for D2, and 3 kcal/mol for SCS and D3); clarify which average is meant.","section":"Supplementary Fig. S1 and main text"},{"comment":"The criteria for classifying modes as 'localized' versus 'delocalized' and for defining the low-frequency 450 THz band are not given; specify how the modes were selected for the 14–16% and >20% contribution estimates.","section":"'Plasmon-like character' section"},{"comment":"The term 'plasmon-like' is used repeatedly without a precise definition of the analogy; a sentence defining the diagnostic (e.g., a participation ratio of the eigenmodes) would help.","section":"Methods and SI"},{"comment":"Minor typographical and clarity issues include the garbled 'semp disp corr .F90' file name and the need to identify in the Fig. 1 caption which curve corresponds to vdW(TS)@SCS.","section":"Methods"}],"recommendation":"major_revision","confidential_remarks":"Confidential to the editor: the manuscript is a methodologically serious paper from a group that developed MBD, and self-citation is understandable given the method basis. The main issues are not novelty but the unverified gas-phase representation of the protein and some overgeneralized wording around cln025. I would not reject on these grounds; a revision with the suggested gas-phase geometry check and quantitative definition of the interaction range could become publishable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nHere is my read on Stöhr and Tkatchenko's paper. The thing to know: the central observation—that many-body dispersion beyond the pairwise approximation destabilizes the gas-phase native state relative to what pairwise models give, and that collective protein-water dispersion compensates this in explicit water—is real and reproducible in the limited but genuine sense that the trend holds across three proteins (Fip35-WW, cln025, HP35-NleNle) and across three pairwise baselines (vdW(TS), D2, D3). The 25 Å range for protein-water correlation in the radial decomposition is the most striking concrete result, and it is a new observation for solvated proteins.\n\nThe paper is honest about most of its own machinery. The MBD parameters and DFTB settings come from earlier benchmarks, not from fitting to the protein folding trajectories. They reuse existing explicit-solvent MD trajectories, which is sensible. The decomposition into self-consistent screening and long-range many-body coupling is clean, and the comparison with pairwise methods is balanced.\n\nThe soft spot is exactly the one the stress-test note flags. The gas-phase leg of the analysis is computed by stripping water from solvated MD snapshots, not by simulating the isolated protein in equilibrium. In vacuo, the protein would relax: side chains pack differently, salt bridges and backbone hydrogen bonds reform, and the hydrophobic core adjusts. The reported 4.5–6 kcal/mol many-body destabilization is on the same order as typical relaxation energies, so the magnitude—and conceivably the sign—of the isolated-protein correction could change if a proper gas-phase ensemble were used. The paper should have flagged this structural mismatch directly; it does not, though the Discussion does acknowledge the limitation of the energetic, fixed-geometry approach.\n\nA second, more minor soft spot is the inherent approximation level of DFTB+MBD. Semi-empirical DFTB and a dipole-based many-body dispersion model cannot yield quantitative free energies, and the authors are appropriately cautious about that. But readers should not take the absolute numbers as predictive.\n\nThe citation pattern is self-referential around MBD, but that is not a defect here because the method is established and the parameters are not fit to the folding data. The leap to long-range biological recognition in the Discussion is speculative, but the authors label it as a broader perspective.\n\nWho is this for? Computational biophysicists and method developers concerned with dispersion in condensed-phase biomolecular modeling. It is not a definitive answer, but it is a thought-provoking data point that will sharpen the debate about pairwise force fields.\n\nMy recommendation: send it to peer review. It deserves a serious referee who can push for a gas-phase relaxation check, or at least an explicit statement that the isolated-protein numbers are for solvated geometries only. I would take the assignment.","headline":"A clear, well-specified computational study showing that many-body dispersion shifts the balance of intra-protein and protein-water energetics; the qualitative claim is credible, but the gas-phase reference is computed on water-stripped solvated snapshots and should be treated as conditional.","tokens_in":16071,"tokens_out":2274,"would_cite":true,"duration_ms":27987,"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":"Many-body van der Waals dispersion forces change which protein states are stable in water, destabilizing the isolated native fold while collective protein-water interactions stabilize it.","keywords":["many-body dispersion","van der Waals interaction","protein folding","protein-water interactions","density-functional tight-binding","collective electronic fluctuations","solvation energy","hydrophobic collapse"],"falsifier":"Recompute the isolated-protein energetics on structures that are genuinely equilibrated in the gas phase, rather than solvent-stripped snapshots from explicit-water trajectories; the claim predicts the MBD-versus-pairwise destabilization of the native state should persist at about 4 to 6 kcal/mol on that ensemble as well. If the sign or magnitude changes materially, the compensation argument in water loses its foundation.","tokens_in":14933,"feed_emoji":"💧","tokens_out":8278,"duration_ms":80654,"temperature":0.7,"pith_summary":"The paper argues that van der Waals dispersion forces in proteins and their aqueous surroundings are quantum-mechanical and many-body in character, and that this character is strong enough to shift the balance of protein folding. Using density-functional tight-binding combined with the many-body dispersion formalism on three fast-folding proteins, it finds that beyond-pairwise dispersion effects destabilize the native state in the gas phase by about 4.5 to 6 kcal/mol relative to pairwise models, because pairwise treatments overcorrelate atoms in the protein core. In explicit water, however, collective protein-water dispersion interactions reverse the tendency and stabilize native conformations and transition states by a comparable amount. The reason is delocalized, plasmon-like electronic fluctuations that couple protein and solvent out to about 25 angstroms. If true, this means pairwise-additive force fields can be fundamentally unbalanced for solvated proteins, and dispersion must be treated explicitly to capture folding energetics.","feed_headline":"Water's collective quantum forces stabilize folded proteins","feed_subtitle":"Many-body dispersion destabilizes isolated folds by ~5 kcal/mol; collective protein-water forces cancel it in water.","key_machinery":"The central machinery is the Many-Body Dispersion (MBD) formalism, a model in which each atom's response is represented by a screened quantum harmonic oscillator and the coupled oscillators are diagonalized to give collective electronic eigenmodes. From the DFTB electron density, effective atomic polarizabilities are obtained, screened self-consistently, and used to build the dipole-coupled Hamiltonian of Equations (2)-(3); the zero-point energy of the coupled oscillators, Equation (5), is the dispersion energy, and a unitary transform generates plasmon-like collective fluctuations. This setup lets the authors decompose the vdW solvation energy into contributions from individual collective modes and compute the radial range of protein-water correlation, which is what reveals the 25 angstrom persistence and the stabilizing role of delocalized electronic motion.","core_discovery":"The paper's central claim is that many-body van der Waals dispersion interactions, not just pairwise additive potentials, play a central role in protein folding energetics in water. Using the MBD formalism, which models electronic fluctuations as coupled quantum harmonic oscillators, the authors find that pairwise treatments such as vdW(TS), D2, and D3 overestimate the stability of the native state of an isolated protein by roughly 4.5 to 6 kcal/mol; about half of this excess comes from neglecting the self-consistent screening of atomic polarizabilities and half from neglecting many-body interactions beyond pairwise terms. When the protein is placed in explicit water, the same many-body treatment shows that the protein-water dispersion interaction acquires a collective, delocalized character that acts oppositely, stabilizing native conformations and transition states by 5 to 7 kcal/mol relative to the unfolded ensemble. The authors trace this stabilization to low-frequency, plasmon-like electronic fluctuations around 450 THz, and show that correlation between protein and solvent remains significant at distances up to 25 angstroms, far beyond the range of pairwise dispersion models.","pith_inferences":["If correlation persists through water at the computed range, the same mechanism should mediate solvent-assisted dispersion interactions between two solutes or between distant parts of a large biomolecule; this is an extension the paper raises but does not compute.","The 450 THz collective modes suggest a concrete spectroscopic test: terahertz-domain probes of solvated proteins should show collective electronic response that tracks folding state, beyond the local water dynamics.","Because MBD softens low-frequency vibrational modes in molecular crystals, the same many-body effects may contribute to folding entropy and free energy, not just internal energy; quantifying entropy is a natural next step.","A testable extension would be to compare MBD-based vdW solvation energies between proteins of different sizes to see whether the pairwise-additivity failure grows with system size, as the 25 angstrom range suggests."],"forward_implications":["Pairwise-additive dispersion models systematically overstabilize the gas-phase native state by about 4.5 to 6 kcal/mol, so gas-phase protein structure and stability predictions require a many-body correction.","Many-body protein-water dispersion stabilizes folded conformations by a comparable amount, so solvation models that treat dispersion pairwise may be missing a folding driving force.","The vdW contribution to the solvation energy drops by 20 to 30 kcal/mol at the hydrophobic collapse, making it a usable folding descriptor even though the total electronic solvation energy is not.","Electronic correlation between protein and solvent is significant up to 25 angstroms, implying that long-range dispersion forces in aqueous biomolecular systems cannot be treated as local.","For the three proteins studied, the effect is independent of secondary structure (beta-sheet, hairpin, and helix), so the conclusion is not specific to one fold type."],"supporting_citations":[{"why":"Supplies the MBD method that computes many-body dispersion beyond pairwise potentials.","marker":"[13]"},{"why":"Establishes the dispersion energy as the zero-point energy of the coupled quantum harmonic oscillators.","marker":"[46]"},{"why":"Provides the DFTB-based scheme for deriving effective atomic polarizabilities used by MBD and vdW(TS).","marker":"[39]"},{"why":"Supplies the density-functional tight-binding electronic structure used for all protein and solvent calculations.","marker":"[40, 41]"},{"why":"Defines the pairwise vdW(TS) baseline that the many-body results are compared against.","marker":"[42]"},{"why":"Supplies the explicit-water folding trajectory of Fip35-WW analyzed throughout the paper.","marker":"[50]"},{"why":"Supplies the explicit-water folding trajectory of cln025 used as a second example.","marker":"[51]"},{"why":"Supplies the explicit-water folding trajectory of HP35-NleNle used as a third example.","marker":"[52, 53]"},{"why":"Provides the collective-mode decomposition used to identify plasmon-like fluctuations and their contribution to solvation energy.","marker":"[18]"},{"why":"Documents the unbalanced pairwise vdW description in molecular mechanics that this paper seeks to explain.","marker":"[37, 38]"}],"fun_headline_variants":["Plasmon-like water interactions tip protein folding balance","Many-body dispersion in water shifts protein stability","Collective quantum fluctuations stabilize proteins in water","Plasmon-like water forces counteract overstabilization of folds","Many-body effects in water change protein folding energetics"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that taking solvated molecular dynamics snapshots and removing the water yields a valid representation of the isolated protein's gas-phase ensemble; if those conformations are not representative, the reported 4.5 to 6 kcal/mol many-body destabilization of the native state could shrink or even reverse.","fun_headline_variants_meta":{"raw":{"variants":["Plasmon-like water interactions tip protein folding balance","Many-body dispersion in water shifts protein stability","Collective quantum fluctuations stabilize proteins in water","Plasmon-like water forces counteract overstabilization of folds","Many-body effects in water change protein folding energetics"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000423,"raw_usage":{"total_tokens":2190,"prompt_tokens":980,"completion_tokens":1210,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":596,"completion_tokens_details":{"reasoning_tokens":1135}},"tokens_in":596,"tokens_out":1210,"duration_ms":9746,"temperature":1.0,"reasoning_tokens":1135,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:52:05.846323+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the isolated-protein energetics on structures that are genuinely equilibrated in the gas phase, rather than solvent-stripped snapshots from explicit-water trajectories; the claim predicts the MBD-versus-pairwise destabilization of the native state should persist at about 4 to 6 kcal/mol on that ensemble as well. If the sign or magnitude changes materially, the compensation argument in water loses its foundation.","supporting_citations":[{"cited_title":"Tkatchenko, R","cited_arxiv_id":null,"evidence_quote":"Supplies the MBD method that computes many-body dispersion beyond pairwise potentials."},{"cited_title":"Tkatchenko, A","cited_arxiv_id":null,"evidence_quote":"Establishes the dispersion energy as the zero-point energy of the coupled quantum harmonic oscillators."},{"cited_title":"St ¨ohr, G","cited_arxiv_id":null,"evidence_quote":"Provides the DFTB-based scheme for deriving effective atomic polarizabilities used by MBD and vdW(TS)."},{"cited_title":"Tkatchenko, M","cited_arxiv_id":null,"evidence_quote":"Defines the pairwise vdW(TS) baseline that the many-body results are compared against."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the explicit-water folding trajectory of Fip35-WW analyzed throughout the paper."},{"cited_title":"Lindorff-Larsen, S","cited_arxiv_id":null,"evidence_quote":"Supplies the explicit-water folding trajectory of cln025 used as a second example."},{"cited_title":"Hermann, D","cited_arxiv_id":null,"evidence_quote":"Provides the collective-mode decomposition used to identify plasmon-like fluctuations and their contribution to solvation energy."}],"review_version":1}