{"id":"b84be084-6e3c-46a5-97ed-587b67573625","arxiv_id":"2411.18931","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Simulating the KcsA selectivity filter with the Orb-D3 neural network potential reveals a T75 hydroxyl-water hydrogen bond that stabilizes water in the filter and enables soft knock-on potassium transport with a conductance estimate of 80 plus or minus 20 pS.","lead":"A neural network trained on crystals was used to simulate the potassium channel's selectivity filter, revealing a previously unseen water-bridging hydrogen bond that may help explain how potassium ions flow. The result tests whether universal machine learning potentials can capture biology they were never trained on.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central mechanism is inferred from a single 5 V forced trajectory, yet the paper itself shows the transport mechanism is force-dependent, so the T75–water hydrogen bond and soft knock-on could be high-voltage artifacts rather than the physiological conduction mechanism.","rationale":"The strongest claim is a mechanistic hypothesis, not an established fact. The paper honestly reports limitations and provides reproducible input files, a video, and a bulk-electrolyte validation of Orb-D3, which are real assets. The T75C mutation is genuine independent evidence that T75 affects K+ conductance, but it does not specifically validate the water-H-bond/co-transport mechanism proposed here. The most vulnerable step is the leap from a 5 V forced, truncated, ~1 ns trajectory to a physiological mechanism, especially because the paper shows that the observed transport mechanism changes with the applied force. The reader's weakest assumption captures this same risk. The proposed force-series test would check whether the T75–water H-bond and soft knock-on survive at near-physiological driving. Until that check is done, the paper supports the observation but not the physiological conclusion, so CONDITIONAL remains the right verdict; my stress-test does not change it.","tokens_in":10433,"tokens_out":6751,"duration_ms":67447,"concrete_test":"Run the identical truncated Orb-D3 system with a series of lower applied forces (0.02, 0.01, and 0.005 eV/A, corresponding to 1, 0.5, and 0.25 V across 50 Å), using metadynamics or umbrella sampling on the SF water/K+ coordinate to overcome the longer timescales. Measure the T75–water H-bond occupancy at S4/S3 and the fraction of conduction events that are soft knock-on. If both decline monotonically with force and extrapolate to zero near the physiological 0.002 eV/A (0.1 V), the proposed mechanism is a high-voltage artifact. If soft knock-on and H-bond occupancy remain robust at the lowest force, the force-driven-artifact concern is refuted.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that the T75 hydroxyl–water hydrogen bond observed with Orb-D3 is a causal stabilizer of water in the selectivity filter (SF) and that soft knock-on with water co-transport is the physiological K+ conduction mechanism. The only direct evidence comes from ~1 ns trajectories of a truncated, frozen-boundary SF driven at 0.1 eV/A, i.e., roughly 5 V across a 50 Å membrane, 25–50 times physiological voltage. Section 3.2 concedes this force corresponds to a transmembrane voltage much higher than physiological voltages, and Section 4 concedes it may induce distortions in the transport mechanism. More importantly, the paper's own results show the mechanism is force-dependent: at forces greater than 0.1 eV/A a hard knock-on mechanism occurs, while at exactly 0.1 eV/A a soft knock-on mechanism emerges. Without a force-dependence study, there is no basis to assume the 0.1 eV/A branch is the physiological one rather than a force-induced crossover. The prior T75C mutation data [8] show that the T75 hydroxyl affects K+ conductance, but they do not show that it acts through water H-bonding or that water co-transport is part of the conductive pathway. The 80 ± 20 pS conductance also averages only the six soft-knock-on runs, excluding the observed hard-knock-on run, and is derived from forced non-equilibrium transit times. Thus the causal chain from H-bond to physiological soft knock-on rests on the unvalidated assumption that the 5 V drive preserves the mechanism.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports ~1 ns molecular dynamics simulations of the selectivity filter (SF) of the KcsA potassium channel using the universal neural network potential Orb-D3-v2. The author observes a hydrogen bond between the T75 side-chain hydroxyl and a water molecule at the SF entrance, and reports a 'soft knock-on' transport mechanism in which water is co-transported with K+ through the SF, with an estimated conductance of 80±20 pS. Additional observations include carbonyl flipping of G77, V76, and T74 residues, and the absence of soft knock-on when the SF is initialized with only K+ ions. The paper argues that the T75 hydroxyl stabilizes water in the SF, enabling selective rapid K+ conduction, and that the T75C mutation experiment provides supporting evidence. The claimed implication is that universal NNPs can reveal biological mechanisms inaccessible to classical force fields.","tokens_in":10658,"tokens_out":5834,"duration_ms":48939,"significance":"If the central claim were established, the paper would be significant in three respects: it would identify a previously unnoticed structural element (the T75–water hydrogen bond) as a key stabilizer of water in the selectivity filter; it would provide a concrete mechanism for water co-transport during K+ conduction, a long-debated question; and it would demonstrate that a universal neural network potential trained on crystals can produce stable, mechanistically informative simulations of a membrane protein. The paper also makes beneficial contributions in sharing input files and analysis scripts, and in including a bulk-electrolyte validation of the potential. However, the current evidence does not yet support the causal claim: the simulations are driven at 0.1 eV/Å (~5 V), the transport mechanism is shown to be force-dependent, the conductance estimate is based on non-equilibrium forced transit times, and the T75C experiment provides only circumstantial support. The central claim therefore remains a plausible hypothesis rather than an established finding, and the paper's abstract and conclusions should be tempered accordingly.","major_comments":[{"comment":"The central claim that the T75–water hydrogen bond enables soft knock-on transport as the physiological mechanism is inferred from simulations driven at 0.1 eV/Å, which the paper itself equates to ~5 V across a 50 Å membrane in Section 3.2. As the author notes in Section 4, this force 'may induce distortions in the transport mechanism.' More importantly, Section 3.1 shows that at forces greater than 0.1 eV/Å a hard knock-on mechanism occurs, while at exactly 0.1 eV/Å soft knock-on appears. This demonstrates that the observed mechanism is force-dependent, and there is no evidence that the 0.1 eV/Å branch is the physiologically relevant one rather than a force-induced crossover. To support the claim, the paper would need either a systematic force-dependence study showing a stable soft-knock-on regime at lower forces, or a free-energy calculation (e.g., PMF) that does not rely on an external bias.","section":"Sections 3.1 and 3.2"},{"comment":"The conductance estimate of 80±20 pS is derived from the average transit time of a full conduction cycle under a 5 V applied force, with the membrane thickness assumed to be 50 Å, and only the six soft-knock-on runs are averaged while the one hard-knock-on run is excluded. This is not a measure of the physiological single-channel conductance, since the applied force drives the system far from equilibrium, and the error bar reflects only the spread between six short trajectories, not statistical or systematic uncertainty. The claim that this value 'falls within the experimental range' in Section 3.2 is therefore misleading; the conductance should be reported as a rough consistency check under non-physiological drive, or removed.","section":"Section 3.2"},{"comment":"The T75C mutation data from Ref. [8] are presented as strong evidence that the T75 hydroxyl group is causally important for potassium conduction. However, that experiment demonstrates only that removing the hydroxyl reduces conductance; it does not show that the hydroxyl acts by hydrogen-bonding to a water molecule in the SF, nor that water co-transport is part of the conductive pathway. The connection between the mutation phenotype and the specific soft knock-on mechanism observed here is an untested inference. The paper should phrase this as a consistency argument rather than as validation of the proposed mechanism.","section":"Section 3.3"},{"comment":"The simulation uses a 20 Å radius cylinder with atoms beyond 15 Å frozen, and Section 4 notes that the SF entrance gradually dehydrates over the course of the simulation due to movements of the surrounding protein and the frozen boundary. Such dehydration could either create or destroy the T75–water hydrogen bond that is central to the proposed mechanism, and no control simulation with a larger or flexible boundary is provided to rule this out. The RDF in Figure 5 is a static structural observable of the T75–T74 interaction, not a direct measure of the water hydrogen bond; the paper would be strengthened by quantifying the lifetime and occupancy of the water–T75 hydrogen bond over the trajectory and correlating it with the conduction events.","section":"Sections 2 and 4"},{"comment":"The observation in Section 3.4 that initializing the SF with four potassium ions and no water produces a hard knock-on mechanism, with the T75 hydroxyl instead hydrogen-bonding to a water molecule outside the SF, shows that the soft knock-on mechanism is not a unique consequence of the T75–water interaction but depends on the initial occupancy and the driving conditions. This is not necessarily a problem, but it should be acknowledged that the soft knock-on mechanism may be one of several conductive states, and the paper's conclusion that T75 'plays a crucial role' in the 'rapid, specific transport' (Section 3.3) is too strong given this variability.","section":"Section 3.4 and Conclusion"}],"minor_comments":[{"comment":"There is a stray 's' at the end of the section; please correct. Additionally, the terminology 'Orb-D3' and 'Orb-D3-v2' is used inconsistently; please settle on one name.","section":"Section 3.5"},{"comment":"The Langevin friction coefficient of 0.01 is given without units; please specify whether this is in units of 1/fs or another unit, and briefly justify the choice.","section":"Section 2"},{"comment":"The caption refers to 'dark blue' for the hydrogen-bond indicator, but the color legend is not defined. Please add a legend or explicitly describe the color coding.","section":"Figure 4"},{"comment":"The statement that NNPs 'cannot handle potassium ions' is too broad; there are NNP models that include K+ in electrolytes. Please qualify the statement to refer to the specific models intended for biological simulations at the time.","section":"Section 1.2"},{"comment":"The average temperature is reported as 307 K despite the thermostat set to 300 K, attributed to 'a small amount of noise on the forces.' A 7 K drift may indicate a thermostat or integration issue; please investigate and report the temperature profile over time, and assess whether this affects the structural conclusions.","section":"Section 2"},{"comment":"The observation of H25 deprotonation is intriguing, but the paper does not discuss whether this could be an artifact of the frozen boundaries or the large applied force. Please add a note on possible artifacts or a control analysis.","section":"Section 3.4"}],"recommendation":"major_revision","confidential_remarks":"The author is affiliated with Orbital Materials and the paper exclusively uses the Orb-D3-v2 potential; this is a potential conflict of interest that should be declared or handled by the editor. The manuscript presents a plausible and clearly described hypothesis, but the evidence in support of the physiological soft knock-on mechanism is insufficient given the non-physiological driving force and the force-dependent crossover demonstrated in the paper itself. The paper would be stronger if the claims were framed as observations under highly non-equilibrium conditions that generate a testable hypothesis, rather than as established mechanisms. The availability of simulation inputs and analysis scripts is a positive aspect that supports reproducibility."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Strictly between us: this paper has a real observation, but the central mechanistic claim is not yet supported. The T75 hydroxyl–water hydrogen bond is directly visible in the trajectories and reproducible across all six simulations that completed a conduction cycle. That part is solid. The paper is also unusually candid about its own limitations, ships input files and analysis scripts, and validates Orb-D3 on bulk KCl, which gives me some confidence the potential is not obviously broken for this system.\n\nThe soft spot is the causal chain from that H-bond to physiological soft knock-on. Every supporting trajectory is driven at 0.1 eV/Å, which corresponds to about 5 V across the membrane. The paper itself says this is much higher than physiological and may distort the transport mechanism. More importantly, the paper shows the mechanism is force-dependent: above 0.1 eV/Å it sees hard knock-on; at exactly 0.1 eV/Å it sees soft knock-on. With no study of how the T75-water bond and the transport mechanism vary with applied force, I cannot tell whether the 0.1 eV/Å branch is the physiological one or a force-induced crossover. The conductance figure is derived from the cycle time under a hand-chosen 5 V drive, and it is averaged over the six soft-knock-on runs, with the one hard-knock-on run excluded. That is a reporting and selection issue in a small sample, even if unintentional. A classical force-field baseline under the same system and protocol is also missing, so the claim that this explains why classical MD underestimates conductance is not directly tested.\n\nThe T75C mutation evidence is real and relevant, but it is an external prior experiment. It shows the hydroxyl matters for conductance; it does not show that it acts through water H-bonding or that water co-transport is the conductive pathway. The paper overreaches slightly when it presents T75C as strong evidence for the specific mechanism.\n\nWho should read this: people interested in potassium channel conduction mechanisms and in testing universal NNPs on proteins. It is a worthwhile pointer, not a settled answer. I would send it to peer review. The right referees will demand a force-dependence study, a PMF or lower-voltage run, a classical baseline, and a report of all runs including the hard-knock-on one. If the soft knock-on branch survives that, it becomes an important result.","headline":"A new, reproducible T75-water H-bond in a KcsA selectivity filter simulation, with an honest but unproven claim that it enables physiological soft knock-on.","tokens_in":11266,"tokens_out":2274,"would_cite":false,"duration_ms":22311,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92C40","92C05"],"pacs":[],"model":"deepseek-v4-flash","headline":"A hydrogen bond between threonine T75 and water in the selectivity filter stabilizes water and enables soft knock-on potassium transport, according to neural-network-potential simulations.","keywords":["potassium channel","KcsA","selectivity filter","neural network potential","molecular dynamics","soft knock-on","water co-transport","hydrogen bond"],"falsifier":"Run the same simulation setup at a physiological driving force (0.2 V or less) with longer simulation time, or with the T75 hydroxyl removed via a T75C mutation: if the soft knock-on water co-transport and the T75–water hydrogen bond no longer appear and conductance drops to the underestimated values of classical force fields, the claim that this bond is the physiological mechanism collapses.","tokens_in":10134,"feed_emoji":"💧","tokens_out":5025,"duration_ms":41364,"temperature":0.7,"pith_summary":"This paper argues that a specific, previously overlooked hydrogen bond is what lets potassium ions flow through the KcsA channel at near-physiological speed. Simulating the selectivity filter with a universal neural network potential, the author observes water molecules being carried through the filter alongside potassium ions — a 'soft knock-on' mechanism that classical force-field simulations have failed to reproduce. The key actor is the hydroxyl side group of the threonine T75 residue, which reaches into the channel entrance and holds a water molecule in place long enough for the next ion to push the whole column forward. If this is right, it would resolve a long-standing debate about whether water co-transport is part of the conduction mechanism, and it would explain why standard simulations underestimate conductance by an order of magnitude. The simulated conductance of 80 ± 20 pS falls inside the experimental range of 40–250 pS.","feed_headline":"A hidden hydrogen bond keeps water flowing through potassium channels","feed_subtitle":"Simulations reveal water hitching a ride through the selectivity filter at 80 ± 20 pS, matching experiment.","key_machinery":"The central objects are the Orb-D3 universal neural network potential — trained on crystal-structure DFT, used here to supply forces for a nanosecond-scale simulation — and the hydrogen bond between the T75 side-chain hydroxyl and a water molecule at the S4 entrance of the selectivity filter. This hydrogen bond is the load-bearing interaction: it anchors water long enough to be co-transported, and its absence in classical force fields is what the paper proposes explains their hard knock-on behavior and low conductances.","core_discovery":"Using the Orb-D3 neural network potential on an 8,450-atom cylinder containing the KcsA selectivity filter, the paper reports a conduction cycle in which a water molecule enters the S4 site, is stabilized by a hydrogen bond between the T75 side-chain hydroxyl and the water, then hops to S3 as potassium ions advance, completing a soft knock-on event. The same mechanism appeared in six independent trajectories, giving a conductance of 80 ± 20 pS that falls inside the experimental 40–250 pS range. The paper further argues that the T75 hydroxyl is causally important, citing the known T75C mutation experiment in which removing this hydroxyl lowers potassium conductance to rubidium-like levels. It also reports water-induced carbonyl flips at G77, V76 and T74 sites that have not all been seen before.","pith_inferences":["A testable prediction follows: if the T75–water hydrogen bond is the stabilizer, then a T75S mutation, which keeps a hydroxyl but changes geometry, should alter but not abolish water co-transport, whereas T75A or T75C should abolish it — distinguishing hydrogen-bond geometry from mere steric effects.","The 5 V driving force may bias the observed mechanism; a stronger test would be enhanced-sampling calculations of the potential of mean force for water entry into S4 with and without the T75 hydroxyl, at zero applied voltage.","The paper's validation of Orb-D3 against a custom NNP for bulk KCl suggests the model's ion-water interactions are reasonable, but the selectivity filter is far outside training data; comparing predicted water occupancy and ion distributions against crystallographic or 2D-IR data would be a more stringent check.","If water co-transport is confirmed, it implies that K+/Na+ selectivity may partly operate through the energy of dehydrating and rehydrating water in the filter, not only through carbonyl coordination of the ions."],"forward_implications":["If water co-transport is real, the selectivity filter is not a purely ionic pore; water moves with ions, affecting osmotic and energetic balances across the membrane.","The T75 hydroxyl becomes a concrete molecular target: mutations that remove or alter this hydrogen bond should reduce conductance, as the T75C mutation already does experimentally.","The observed carbonyl flips at G77, V76 and T74 may explain low S2 ion occupancy and could be linked to c-type inactivation, connecting the conduction mechanism to channel gating.","Universal neural network potentials trained on crystal data can capture flexible protein dynamics that classical force fields miss, opening membrane-protein simulation to these tools.","A conductance within the experimental range supports the soft knock-on mechanism as the physiological one, rather than the hard knock-on seen in classical molecular dynamics."],"supporting_citations":[{"why":"Provides the crystallographic picture of the selectivity filter with alternating ions and water molecules that the simulations start from.","marker":"[2]"},{"why":"Source of the initial simulation structures and coordinates for the selectivity-filter system.","marker":"[19]"},{"why":"The T75C mutagenesis experiment showing hydroxyl removal collapses potassium conductance; key causal evidence for the paper's mechanism.","marker":"[8]"},{"why":"Experimental conductance range of 40–250 pS used to judge the simulated 80 ± 20 pS as plausible.","marker":"[44]"},{"why":"The Orb-D3 universal neural network potential used for all simulations in the paper.","marker":"[35]"},{"why":"Reduced electron density at the S4 site upon hydroxyl removal and rubidium behavior, used to interpret the destabilizing effect of removing T75's hydroxyl.","marker":"[45]"},{"why":"Custom-trained neural network potential simulations of bulk KCl whose RDFs are used to validate Orb-D3's potassium-water interactions.","marker":"[52]"}],"fun_headline_variants":["Water co-transport revealed in potassium channels by neural network","Hidden hydrogen bond enables water's soft knock-on in K+ channels","Neural network reveals hidden water bridge in potassium channel","Water rides along: soft knock-on in potassium channels"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The 0.1 eV/Å driving force (about 5 V across the membrane) plus the frozen outer cylinder of protein is assumed to reproduce the physiological conduction mechanism; the paper itself notes this force is far above physiological voltages and may distort the mechanism.","fun_headline_variants_meta":{"raw":{"variants":["Water co-transport revealed in potassium channels by neural network","Hidden hydrogen bond enables water's soft knock-on in K+ channels","Neural network reveals hidden water bridge in potassium channel","Water rides along: soft knock-on in potassium channels"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001544,"raw_usage":{"total_tokens":6150,"prompt_tokens":898,"completion_tokens":5252,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":514,"completion_tokens_details":{"reasoning_tokens":5185}},"tokens_in":514,"tokens_out":5252,"duration_ms":34733,"temperature":1.0,"reasoning_tokens":5185,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T10:44:19.804597+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same simulation setup at a physiological driving force (0.2 V or less) with longer simulation time, or with the T75 hydroxyl removed via a T75C mutation: if the soft knock-on water co-transport and the T75–water hydrogen bond no longer appear and conductance drops to the underestimated values of classical force fields, the claim that this bond is the physiological mechanism collapses.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Source of the initial simulation structures and coordinates for the selectivity-filter system."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the crystallographic picture of the selectivity filter with alternating ions and water molecules that the simulations start from."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The T75C mutagenesis experiment showing hydroxyl removal collapses potassium conductance; key causal evidence for the paper's mechanism."},{"cited_title":"LeMasurier, L","cited_arxiv_id":null,"evidence_quote":"Experimental conductance range of 40–250 pS used to judge the simulated 80 ± 20 pS as plausible."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reduced electron density at the S4 site upon hydroxyl removal and rubidium behavior, used to interpret the destabilizing effect of removing T75's hydroxyl."}],"review_version":1}