{"id":"4684250c-e5ac-4f23-a809-81e2db918c9a","arxiv_id":"2506.20205","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A SASA plus ReaxFF screening approach predicts plasma-induced amino acid modifications, with phenylalanine oxidation in CviUPO confirmed by mass spectrometry.","lead":"This preprint combines solvent accessibility analysis and reactive molecular dynamics to predict which amino acids in three enzymes are attacked by plasma-generated reactive species, then tests a subset of predictions on one enzyme with mass spectrometry. The method correctly flagged phenylalanine oxidation in a plasma-treated peroxygenase, but the confirmation is partly circular because the prediction prompted the search for that exact modification.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The alleged MS confirmation is not independent: Phe oxidation is a known plasma effect and the search was seeded by the prediction, so the current data cannot support the central claim.","rationale":"The most load-bearing part of the paper is not only the force-field accuracy but the evidential value of the MS confirmation. The authors chose to search for Phe modification because the SASA screen indicated Phe, and Phe oxidation is known to occur in plasma-treated proteins (refs. 43,44); finding Phe oxidation is therefore weak evidence that the screen has predictive power. The absence of predicted Lys/Arg modifications and the unanticipated Met oxidation further weaken the discrimination. The SI validation for OH at 300 K did not rank Phe positions among the top ten interaction points, so the residue-specific claim in the strongest_claim is not actually shown by the manuscript. The SI caveat about ReaxFF radical/ion indistinguishability is a real limitation, but it is explicitly acknowledged by the authors and mainly affects mechanistic attribution; the non-independent validation affects the central claim directly. A blinded, pre-registered validation on a second enzyme with a broad modification panel and an enrichment test would settle this. Since such a test is feasible and the computational method may still be useful, the conditional verdict remains appropriate; I would not escalate to reject, but the confirmation claim should be downgraded until an independent validation is performed.","tokens_in":21266,"tokens_out":8847,"duration_ms":104444,"concrete_test":"Reanalyze the raw MS data (or, if unavailable, require deposition) with a broad, pre-specified variable-modification panel that includes Phe+O, Tyr+O, Trp+O, His+O, Met+O, Cys oxidations, and Lys/Arg dehydrogenation, without using the SASA prediction to choose the searched modifications. Then test whether the SASA-predicted Phe residue types or positions are statistically enriched among the confirmed modified residues relative to all candidate residues (e.g., Fisher exact test on residue counts, with the 30-point cutoff removed or justified). If Phe oxidation appears without SASA seeding and the enrichment is not significant, the current 'confirmation' does not support the central claim; if the enrichment is significant, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that SASA/ReaxFF screening predicted phenylalanine modification and mass spectrometry confirmed it—rests on a validation that is neither independent nor discriminating. The Conclusion states that 'the Phe modification by plasma treatment has thus been brought into focus by the SASA analysis, which prompted the inclusion of Phe modifications in the MS database search.' Because the MS search was seeded by the prediction, the same data cannot serve as independent confirmation; the discovery and validation sets are not separated. Phe hydroxylation is a well-documented plasma-induced modification (refs. 43,44) and would be expected even if the SASA screen carried no predictive information. The observed Met42/Met220 oxidation was not predicted, while predicted Lys/Arg interactions were not observed as modifications, and the SI validation table for OH at 300 K lists Leu225, Asp91, Arg84, and Val153 as top interaction points, not Phe51/67/208. Thus the experimental evidence supports only a generic type-level sensitivity of Phe, not the residue-specific or method-specific predictive power claimed. With one enzyme, one condition, and one positive modification among several searched, the MS data cannot distinguish the method's prediction from a generic ROS background. The SI's own caveat about ReaxFF's inability to distinguish OH radicals from OH ions ('Explanation to the species nature in ReaxFF') further weakens the computational half of the argument, but the non-independent validation is the more load-bearing problem.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a computational workflow to predict which amino-acid residues of enzymes are modified by plasma-generated species (PGS) such as H2O2, OH, O2, O, NO, and H. The workflow combines a SASA-based interaction analysis, in which a single probe species is placed at solvent-accessible surface points and ReaxFF interaction energies are computed (Eq. 1), with reactive molecular dynamics (ReaxFF with the Monti et al. force field) as a short-time validation of bond formation at predicted positions. The method is applied to three enzymes, with CviUPO as the main experimental subject. For CviUPO, the analysis predicts strong OH interactions with Lys, Phe, and Arg; a DBD plasma-treated sample is then analyzed by LC-MS/MS, and the authors report significant oxidation of Phe51, Phe67, Phe208, Met42, and Met220, claiming that the predicted modification of phenylalanine was experimentally confirmed. The Conclusion and the Supplementary Information explicitly acknowledge that the Phe prediction prompted the inclusion of Phe modifications in the MS database search and that ReaxFF cannot reliably distinguish OH radicals from OH ions.","tokens_in":21465,"tokens_out":9766,"duration_ms":101096,"significance":"If validated, the proposed screening workflow would be practically useful for the plasma-biocatalysis community as a cheap computational prescreen for ranking amino-acid residues on a protein surface with respect to reactivity toward plasma-generated species, and for directing time-consuming mass spectrometry searches. The manuscript has notable strengths: the SASA-analysis package is released on GitHub, the simulation and mass spectrometry data are provided, and the SI candidly documents the ReaxFF species limitation. However, the current evidence does not establish the central claim as stated. The MS validation is partly circular because the search space was seeded by the prediction; Phe oxidation is a known general plasma effect; the agreement is at the level of residue type rather than residue identity; and the force-field ambiguity acknowledged in the SI affects the entire interaction-energy ranking. The paper currently supports a screening heuristic with a plausible consistency check, not an independently confirmed predictive method.","major_comments":[{"comment":"The Conclusion states that 'the Phe modification by plasma treatment has thus been brought into focus by the SASA analysis, which prompted the inclusion of Phe modifications in the MS database search.' Because the prediction seeded the database search, the subsequent detection of Phe oxidation is not an independent confirmation of the method; it is a consistency check whose search space was conditioned on the prediction. This circularity is compounded by the fact that Phe hydroxylation is a well-documented plasma-induced modification of amino acids in the very literature cited by the authors (refs. 43 and 44), so its observation would be expected even if the SASA screen carried no predictive information. The Abstract's claim that predictions were 'confirmed experimentally by mass spectrometry' should be weakened to a consistency statement, or the authors should provide a discriminating validation, e.g., a prospective or pre-registered MS search over a fixed modification list in which the Phe prediction was made before the search was executed.","section":"Conclusion"},{"comment":"The experimental match is at the amino-acid-type level, not the residue level. The ten strongest OH interaction points at 300 K listed in Supplementary Table 1 map to Leu225, Asp91, Arg84, Val153, and other non-Phe residues, and neither the short reactive-MD validation nor the high-concentration simulations (Fig. 5) report bond formation at Phe51, Phe67, or Phe208; they report Phe only as a residue type. The SASA evidence for Phe comes from the frequency-normalized counts in Fig. 2(e), i.e., Phe is over-represented in the pool of strong-interaction points, not from residue-specific interaction energies at the three positions later observed by MS. The paper should explicitly state that the method predicts which amino-acid types are preferential targets and that residue-level correspondence between prediction and experiment is not currently demonstrated.","section":"Results, SASA interaction analysis (CviUPO); Supplementary Table 1"},{"comment":"The SI states that distinguishing OH radicals from OH ions in ReaxFF 'remains highly uncertain' and that 'any conclusions drawn from these simulations should be interpreted with caution,' and the main text acknowledges that the modeled 'rmdPGS ... are not real plasma-generated species.' Since the SASA interaction energies (Eq. 1) and all bond-formation statistics in the reactive MD are computed with this force field, the mechanistic attribution of Phe oxidation to OH and O in the Conclusion is not supported by a species-resolved model. This caveat should be moved into the main text and the mechanistic claims reframed as applying to the modeled species only. A concrete strengthening would be to benchmark the force field against an experimentally known relative reactivity order, e.g., the high susceptibility of Met and Cys reported in refs. 43-45, before using it to rationalize Phe-selective modification.","section":"SI, 'Explanation to the species nature in ReaxFF'"},{"comment":"The screen produced a false negative for a substantial fraction of the experimentally observed modifications: the SASA analysis reports no interactions with Met or Cys because these residues are rare and buried, yet Met42 and Met220 are oxidized in all three replicates. The paper's explanation that buried residues become accessible post hoc does not rescue the predictive claim, since the same argument could apply to many buried residues and is not testable from the reported data. The authors should quantify the screen's miss rate (two of the five observed modified residues belong to a residue type the screen excluded) and discuss whether the Met oxidation indicates that the static SASA screen misses modifications that require protein dynamics, radical diffusion into the protein interior, or species not represented in the model. The weak experimental signal for the strongly predicted Lys/Arg modifications (detected only in some replicates and at low relative intensity) further limits what the current experiment can establish about the method's selectivity.","section":"Results, Mass spectrometry of plasma-treated CviUPO"},{"comment":"The mass spectrometry table reports modified Phe51, Phe67, and Phe208, with Phe51 and Phe67 located in the same 33-residue tryptic peptide, but no site-localization evidence (e.g., fragment-ion coverage or a localization score) is provided, and no error bars or significance tests are given for the relative intensities. Because the headline claim is residue-specific oxidation, the authors should report the localization confidence for each modified site (e.g., the number of site-determining fragment ions) and descriptive statistics over the three replicates, and define the 'relative intensity > 100' acceptance criterion.","section":"Table 1"}],"minor_comments":[{"comment":"The figure caption is internally inconsistent: it is titled 'Bond analysis for high concentrations of hydrogen' while the surrounding text and the axis description refer to OH, and the caption itself mentions 'an additional OH atom' while the panels are described in terms of hydrogen. Please reconcile the caption with the plotted species.","section":"Figure 5"},{"comment":"References 39 and 43 appear to cite the same article by Takai et al. (2014) with different journal names (Journal of Physics D: Applied Physics vs. Plasma Sources Science and Technology); please deduplicate and correct.","section":"References"},{"comment":"In Eqs. (2) and (3), the quantity 'Interactions Total Number' is not defined; please specify whether it is the total number of SASA points, the number of points above the top-30 cutoff, or the summed interaction counts over the analyzed set.","section":"Computational details"},{"comment":"The default cutoff of the thirty strongest interaction energies is described only as 'determined empirically' with no supporting data; a brief sensitivity check (e.g., top 10 vs. top 50) would help the reader judge the robustness of the residue rankings and is recommended given the load this parameter carries.","section":"Computational details"},{"comment":"The acceptance criterion 'relative intensity > 100' is undocumented; please state how the relative intensities were computed (e.g., normalization to peptide or protein abundance) and why the threshold of 100 was chosen.","section":"Table 1"},{"comment":"Several typos and LaTeX artifacts remain in the text, including 'occurring in textitCviUPO', the broken 'N V Ten' ensemble label, and the phrase 'are display in Figure 5'; a careful proofread is needed.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is unusually transparent about its own limitations: the SI explicitly disclaims the ReaxFF species assignment, and the Conclusion openly describes how the Phe prediction seeded the MS search. That transparency is commendable, but the abstract and headline framing ('confirmed experimentally') overstate what the evidence supports. I do not see grounds for integrity concerns; the issues are framing, validation design, and the mismatch between type-level and residue-level claims, all of which are addressable within the manuscript's scope. The SASA-analysis code release on GitHub is a genuine strength that should be preserved. The paper fits the journal's scope as a methods-oriented bio-physics contribution. I would recommend major revision with emphasis on (1) reframing the validation as a consistency check or providing an independent test, (2) distinguishing type-level from residue-level predictions, and (3) moving the force-field caveat into the main text."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Adrian,\n\nQuick take: this paper is worth knowing about, but the headline claim overreaches. The authors built a cheap screening pipeline—SASA interaction energies plus short ReaxFF MD—to predict which amino-acid types on an enzyme surface are likely to react with plasma-generated species. They applied it to CviUPO, AaeUPO, and GapA, and they provide new MS data for plasma-treated CviUPO. The software is on GitHub, the methods are described in enough detail to reproduce, and the SI is refreshingly candid about ReaxFF's inability to distinguish OH radicals from OH ions. That honesty is real credit.\n\nWhat's actually new: applying this SASA-plus-ReaxFF combination to predict plasma-induced protein modification targets, and generating the CviUPO modification dataset. As a pre-screen to direct MS experiments, the idea has legs.\n\nBut the main claim—that the prediction of phenylalanine modification was confirmed by MS—does not hold up. The confirmation is not independent: the SASA analysis prompted including Phe in the MS search, and Phe oxidation is already a documented plasma effect (their own refs. 43, 44). More telling, the residue-level predictions don't match the experiment. The ten lowest-energy OH SASA points for CviUPO at 300 K are Leu225, Asp91, Arg84, Val153 (SI Table 1), not Phe51, Phe67, or Phe208. The high-concentration MD shows Phe as a frequent target at the amino-acid-type level, but not those specific residues. So the method did not predict the observed oxidation sites. At the same time, Met42 and Met220 were oxidized even though SASA/ReaxFF predicted essentially no Met or Cys interactions. That's a counterexample sitting in the same table.\n\nThere are also smaller issues: the 30-point cutoff for relative interactions is empirical and not justified, and the MS data are not deposited. The ReaxFF species-ambiguity caveat is acknowledged, but it remains load-bearing for any specific interaction-energy ranking.\n\nWho should read this: plasma biocatalysis researchers, and computational people working on protein-surface reactivity screens. It deserves serious peer review—the workflow is publishable—but not in its current form. The authors need to separate discovery from validation, deposit data, and test on a second enzyme where the MS search is not seeded by the prediction. If the method survives that, it becomes genuinely useful. As written, it's an interesting proof-of-concept with an overinterpreted confirmation.\n\nMy recommendation: send it to review with a major-revision request. The core idea should not be desk-rejected, but the validation logic must be fixed.\n\nBest,\n[Your name]","headline":"Useful screening workflow with an honest limitations section, but the central validation claim is not supported: the MS search was seeded by the prediction, and the residue-level top hits from SASA were not the experimentally modified sites.","tokens_in":22060,"tokens_out":2576,"would_cite":false,"duration_ms":31845,"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":"A solvent-accessible-surface screen predicts which amino acid residues plasma species modify, and mass spectrometry confirms three phenylalanine oxidations on CviUPO.","keywords":["solvent accessible surface area","plasma-generated species","reactive molecular dynamics","phenylalanine oxidation","unspecific peroxygenase","plasma-driven biocatalysis","protein modification screening","mass spectrometry"],"falsifier":"Plasma-treat CviUPO with and without a hydroxyl-radical scavenger and compare oxidation of Phe51, Phe67, and Phe208 by mass spectrometry; if oxidation persists when OH radicals are scavenged, the claim that OH drives these modifications is wrong. A complementary check would compute the OH–phenylalanine reaction energy with a high-level quantum-chemical method on a model compound and compare it with the ReaxFF value used in the screening.","tokens_in":21041,"feed_emoji":"⚡","tokens_out":9699,"duration_ms":97930,"temperature":0.7,"pith_summary":"Plasma-driven biocatalysis uses a discharge to supply H2O2 to enzymes, but the same plasma generates radicals that can oxidize the enzyme. This paper argues that a solvent-accessible-surface-area (SASA) interaction screen, combined with reactive molecular dynamics, can identify amino acid residues most likely to be modified by plasma-generated species. Applied to the peroxygenase CviUPO, the screen flagged phenylalanine and basic residues as OH interaction partners, and mass spectrometry confirmed oxidation of Phe51, Phe67, and Phe208 after plasma treatment, together with two methionines that the screen had not singled out. The paper's conclusion is that this SASA-based screen is a fast method for directing targeted mass spectrometry searches and for locating regions of a protein surface that small reactive molecules will attack.","feed_headline":"Surface map predicts which protein spots plasma oxidizes","feed_subtitle":"For the enzyme CviUPO, the screen flags phenylalanine; mass spectrometry confirms three oxidized sites.","key_machinery":"The machinery is a two-stage computational pipeline. First, a SASA interaction analysis places a probe molecule at each solvent-accessible point about 1.4 Å from the protein surface, rotates it to the best orientation, and computes a local interaction energy using the ReaxFF reactive force field, a bond-order potential that allows bonds to break and form during the simulation; post-processing ranks residues by total and frequency-normalized (relative) interactions, using the thirty strongest energy points. This yields a three-dimensional interaction map for each plasma species. Second, reactive molecular dynamics validate the hot spots: short 75 fs simulations at the ten lowest-energy positions test whether a covalent bond forms, and longer high-concentration simulations with randomly placed species check whether the residue preferences survive diffusion and competition with solvent. The pipeline was run on three enzymes, CviUPO, AaeUPO, and GapA, and experimentally checked on CviUPO.","core_discovery":"The central claim is that local interaction energies computed at solvent-accessible points on a protein surface identify the residues most likely to be chemically modified by reactive plasma species. The method places one model plasma species at a time on surface points 1.4 Å from the protein, relaxes its orientation, and computes a local interaction energy $E_{\\mathrm{int}} = E_{\\mathrm{SASA}} - E_{\\mathrm{macro\\,mol}} - E_{\\mathrm{probe\\,mol}}$ from the ReaxFF reactive force field. Hot spots cluster on specific residues; reactive molecular dynamics at the ten lowest-energy sites shows that H, O, and OH form bonds there, and high-concentration simulations reproduce the same residue preferences even with random starting positions. For CviUPO, the screen predicted strong interactions with surface-exposed phenylalanines, and dielectric-barrier-discharge plasma treatment followed by tryptic digestion and LC-MS/MS detected significant oxidation of Phe51, Phe67, and Phe208 in all replicates. The paper therefore claims that SASA-based screening can direct targeted mass spectrometry searches and, more generally, locate protein-surface regions of interest for small reactive molecules.","pith_inferences":["A testable extension beyond this paper: benchmark the SASA screen on enzymes with already published plasma- or oxidant-induced modification maps, asking whether the rank order of predicted hot spots matches experimentally observed modified residues.","Editorial inference: the screen's residue preferences may be driven mostly by surface exposure and electrostatics rather than by accurate radical chemistry; comparing SASA hot spots against a surface-exposure-only ranking would show how much the reactive dynamics add.","If Phe oxidation near the substrate channel indeed impairs catalysis, engineering solvent-exposed phenylalanines into less oxidizable residues could improve enzyme stability under plasma-driven conditions.","The current MS search covered only a small panel of modifications, so the actual damage may be broader; a wider search including nitration and carbonyl formation could test whether other SASA-flagged residues, such as lysine and arginine, are modified under conditions that favor reactive nitrogen species."],"forward_implications":["If the method is right, any enzyme can be screened before plasma experiments to obtain a short list of residues to search for modification in MS data, reducing the search space for time-consuming proteomics.","For CviUPO, the oxidation of Phe51 and Phe67, both near the substrate channel, flags surface-exposed aromatic residues as a potential cause of activity loss under plasma treatment.","The solvent simulations say most reactive species react with water before reaching the enzyme; this directly supports immobilization or geometry changes that lengthen the diffusion path as protection strategies.","At elevated temperatures, unfolding exposes buried residues and the heme cofactor to attack, so thermal control of the reaction environment should limit damage to catalytically essential regions.","Because the interaction energy definition is not specific to plasma species, the same SASA screen could be used to locate binding or modification sites for other small reactive molecules."],"supporting_citations":[{"why":"Supplies the biomolecule ReaxFF force field used for all reactive molecular dynamics and SASA interaction energies.","marker":"[33]"},{"why":"Defines the ReaxFF bond-order formalism that allows bonds to break and form in the simulations.","marker":"[32]"},{"why":"Provides the molecular graphics engine used to determine solvent-accessible surface points.","marker":"[34]"},{"why":"Establishes the 1.4 Å probe-distance convention for solvent-accessible surface calculations.","marker":"[35]"},{"why":"Introduces plasma-driven in situ H2O2 production for biocatalysis, the application context the study addresses.","marker":"[9]"},{"why":"Describes the CviUPO catalytic mechanism and identifies active-site residues Glu162 and Lys165 used to interpret heme-channel damage.","marker":"[42]"},{"why":"Documents plasma-induced oxidation of free phenylalanine, supporting the plausibility of the observed Phe hydroxylation.","marker":"[43]"},{"why":"Shows that prolonged diffusion through solvent reduces enzyme modification by plasma species, matching the solvent simulation trend.","marker":"[46]"}],"fun_headline_variants":["Surface energy map predicts plasma oxidation spots on proteins","Phenylalanine hotspots on enzymes flagged by surface scoring","Mass spec confirms phenylalanine sites predicted by plasma screen","Local interaction energies find protein residues plasma modifies","Surface accessibility screen predicts plasma protein oxidation"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the reactive force field faithfully represents the chemistry of plasma species such as hydroxyl radicals, and the paper's own supporting information concedes that distinguishing OH radicals from OH ions in ReaxFF 'remains highly uncertain'; if that chemistry is misrepresented, the predicted interaction profiles and modification sites may not reflect real plasma conditions.","fun_headline_variants_meta":{"raw":{"variants":["Surface energy map predicts plasma oxidation spots on proteins","Phenylalanine hotspots on enzymes flagged by surface scoring","Mass spec confirms phenylalanine sites predicted by plasma screen","Local interaction energies find protein residues plasma modifies","Surface accessibility screen predicts plasma protein oxidation"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0004,"raw_usage":{"total_tokens":2121,"prompt_tokens":1010,"completion_tokens":1111,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":626,"completion_tokens_details":{"reasoning_tokens":1041}},"tokens_in":626,"tokens_out":1111,"duration_ms":19362,"temperature":1.0,"reasoning_tokens":1041,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T22:53:14.989834+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Plasma-treat CviUPO with and without a hydroxyl-radical scavenger and compare oxidation of Phe51, Phe67, and Phe208 by mass spectrometry; if oxidation persists when OH radicals are scavenged, the claim that OH drives these modifications is wrong. A complementary check would compute the OH–phenylalanine reaction energy with a high-level quantum-chemical method on a model compound and compare it with the ReaxFF value used in the screening.","supporting_citations":[{"cited_title":"VMD -- V isual M olecular D ynamics","cited_arxiv_id":null,"evidence_quote":"Supplies the biomolecule ReaxFF force field used for all reactive molecular dynamics and SASA interaction energies."},{"cited_title":"L.; Shin, Y","cited_arxiv_id":null,"evidence_quote":"Defines the ReaxFF bond-order formalism that allows bonds to break and form in the simulations."},{"cited_title":"P.; Wright, W","cited_arxiv_id":null,"evidence_quote":"Provides the molecular graphics engine used to determine solvent-accessible surface points."},{"cited_title":"Visualization and analysis of atomistic simulation data with OVITO –the Open Visualization Tool","cited_arxiv_id":null,"evidence_quote":"Establishes the 1.4 Å probe-distance convention for solvent-accessible surface calculations."},{"cited_title":"F.; Hollmann, F.; Alcalde, M.; Kourist, R.; Bandow, J","cited_arxiv_id":null,"evidence_quote":"Introduces plasma-driven in situ H2O2 production for biocatalysis, the application context the study addresses."},{"cited_title":"Chemical modification of amino acids by atmospheric-pressure cold plasma in aqueous solution","cited_arxiv_id":null,"evidence_quote":"Describes the CviUPO catalytic mechanism and identifies active-site residues Glu162 and Lys165 used to interpret heme-channel damage."},{"cited_title":"E.-D.; Aadil, R","cited_arxiv_id":null,"evidence_quote":"Documents plasma-induced oxidation of free phenylalanine, supporting the plausibility of the observed Phe hydroxylation."},{"cited_title":"J.; Van Genechten, K.; Gasteiger, J","cited_arxiv_id":null,"evidence_quote":"Shows that prolonged diffusion through solvent reduces enzyme modification by plasma species, matching the solvent simulation trend."}],"review_version":1}