arxiv: 2604.14906 · v2 · submitted 2026-04-16 · ⚛️ physics.bio-ph · cs.LG

Recognition: unknown

Unraveling the Mechanism of Drug Binding to SARS-CoV-2 RNA Pseudoknot with Thermodynamics-Driven Machine Learning

Mariia Ivonina , Jakub Rydzewski

Authors on Pith no claims yet

Pith reviewed 2026-05-10 09:59 UTC · model grok-4.3

classification ⚛️ physics.bio-ph cs.LG

keywords SARS-CoV-2RNA pseudoknotprogrammed ribosomal frameshiftingmolecular dynamicsmachine learningfree energy landscapeligand bindingprotonation state

0 comments

The pith

Ligand-induced destabilization of the SARS-CoV-2 RNA pseudoknot is selective to its topology and the ligand's protonation state.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper applies a machine learning method called spectral map to molecular dynamics simulations to identify the slow collective motions in the RNA pseudoknot when bound to merafloxacin and analogs. It finds that the ligands destabilize different helical stems depending on whether the pseudoknot is threaded or unthreaded. The zwitterionic form of the ligand adds slow dynamics to the unthreaded form, unlike the neutral form. This shows that both the RNA fold and the drug's charge matter for how it alters the structure relevant to viral frameshifting.

Core claim

Using spectral map on MD trajectories, the free-energy landscapes demonstrate that merafloxacin and analogs destabilize the S2 stem in the threaded pseudoknot but shift destabilization to S1 and S3 stems in the unthreaded pseudoknot, with the zwitterionic form uniquely imposing slow dynamics on the unthreaded topology, and neutral and zwitterionic forms differing in mechanisms within the same topology.

What carries the argument

Spectral map, a thermodynamics-driven machine learning method that learns collective variables isolating the slowest dynamic modes from MD trajectories of the RNA-ligand complexes.

If this is right

Physiological protonation states must be considered when modeling RNA-targeted antiviral action.
Drug design for -1 PRF inhibitors should account for pseudoknot topology to achieve specific destabilization.
The method can distinguish ligand effects on different RNA conformations.
Neutral and ionized ligand forms can have qualitatively different mechanisms on the same RNA.
The unthreaded pseudoknot becomes dynamically active only with certain ligand forms.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

These findings suggest that targeting specific topologies could improve selectivity of antivirals against SARS-CoV-2.
Extending the approach to other viral RNAs or ligands might reveal common patterns in frameshifting regulation.
Experimental validation could involve mutating stems to see if ligand effects change as predicted.
Protonation-dependent dynamics imply pH could modulate drug efficacy in cellular environments.

Load-bearing premise

The spectral map method correctly identifies the slowest relevant dynamic modes in the MD simulations without being affected by force field inaccuracies or fast fluctuations.

What would settle it

If experiments or higher-level simulations show that the S2 stem remains stable in the threaded form upon ligand binding, or that the zwitterionic form does not slow the unthreaded pseudoknot, the topology-selective destabilization claim would be falsified.

Figures

Figures reproduced from arXiv: 2604.14906 by Jakub Rydzewski, Mariia Ivonina.

**Figure 2.** Figure 2: Secondary structures of the SARS-CoV-2 −1 ribosomal frameshift-stimulatory pseudoknot derived from crystallographic data. (a) Threaded (RT) conformation, 9 in which the 5′ -end passes through the junction formed by stems 1 (S1), 2 (S2), and 3 (S3), thereby producing a topological tertiary fold. (b) 3D representation of the RT model. (c) Unthreaded (RU) conformation, 11 in which this topological fold is abs… view at source ↗

**Figure 3.** Figure 3: Ligands considered in this study: (left) Merafloxacin in the neutral (MERA [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗

**Figure 4.** Figure 4: Comparison of structural stability and conformational dynamics of the ligand [PITH_FULL_IMAGE:figures/full_fig_p017_4.png] view at source ↗

**Figure 5.** Figure 5: Structural stability and conformational dynamics of the threaded (RT) topology of [PITH_FULL_IMAGE:figures/full_fig_p019_5.png] view at source ↗

**Figure 6.** Figure 6: Structural stability and conformational dynamics of the unthreaded (RU) topology [PITH_FULL_IMAGE:figures/full_fig_p020_6.png] view at source ↗

**Figure 7.** Figure 7: Eigenvalues λk corresponding to the converged CVs after SM learning. Dots and error bars show means and standard deviations, respectively, calculated over the last 50 epochs. Insets show the training loss L during 100 epochs. 4.3 Slow Modes Distinguish Threaded and Unthreaded Topologies The eigenspectra of the converged CVs reveal a fundamental dynamical distinction between the threaded and unthreaded pseu… view at source ↗

**Figure 8.** Figure 8: Free-energy landscapes F(z) reconstructed in CV space z = (z1, z2) learned by SM for each system. Each state is labeled by S0, S1, and S2. The rotation of each landscape is set so that the state labels are ordered clockwise. The ensembles corresponding to the metastable states depict 10 structures closest to the minima (white dots). 23 [PITH_FULL_IMAGE:figures/full_fig_p023_8.png] view at source ↗

read the original abstract

The SARS-CoV-2 RNA pseudoknot is a promising target for antiviral intervention, as it regulates the efficiency of $-$1 programmed ribosomal frameshifting ($-$1 PRF), a mechanism that is essential for viral protein synthesis. The pseudoknot represents a viral RNA sequence composed of helical stems that adopts two long-lived topologies, threaded and unthreaded. Ligand-induced distortion of this fold is thought to underlie the susceptibility of $-$1 PRF to small-molecule inhibitors. Resolving these distortions from unbiased molecular dynamics (MD) requires collective variables (CVs) that isolate the slowest dynamic modes of the RNA--ligand system from the high-frequency fluctuations. Here, we use spectral map (SM), a thermodynamics-driven machine-learning method, to learn such CVs directly from MD trajectories of the SARS-CoV-2 RNA pseudoknot in complex with the $-$1 PRF inhibitor merafloxacin and two related analogs. We examine both threaded and unthreaded pseudoknot topologies and consider the neutral and ionized ligand forms relevant at physiological pH. Free-energy landscapes show that ligand-induced destabilization is topology-selective: merafloxacin and its analogs destabilize the S2 stem in the threaded pseudoknot, whereas in the unthreaded pseudoknot, destabilization shifts to the S1 and S3 stems. We find that the zwitterionic form of merafloxacin uniquely imposes slow dynamics on the otherwise featureless unthreaded pseudoknot. Furthermore, the neutral and zwitterionic forms of merafloxacin differ qualitatively in their mechanisms within the same RNA topology. Overall, these results clarify how pseudoknot topology, ligand type, and protonation state shape the slow conformational dynamics of viral RNA and establish physiological protonation as an essential factor for modeling RNA-targeted drug action.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper finds that merafloxacin destabilizes different stems in the SARS-CoV-2 pseudoknot depending on topology and protonation state, but the spectral map results lack reported checks against sampling or force-field issues.

read the letter

The main thing to know is that this work uses spectral mapping on MD trajectories to show ligand-induced stem destabilization is selective: S2 in the threaded pseudoknot, S1 and S3 in the unthreaded one, with the zwitterionic merafloxacin adding slow dynamics only in the unthreaded case. Neutral and charged forms also act differently within the same topology. These observations on protonation dependence and topology selectivity are the concrete new pieces, and they line up with the goal of understanding how small molecules might tune -1 PRF efficiency. The approach of pulling CVs directly from the trajectories across multiple conditions is a reasonable way to handle the high-frequency noise in RNA simulations. Considering physiological protonation states is a clear plus, since it avoids an obvious modeling shortcut. The soft spot is exactly the one flagged in the stress test. The free-energy landscapes come from learned spectral map variables, yet the abstract and available details give no convergence diagnostics on the eigenvalues, no side-by-side comparison with ordinary stem RMSD or base-pair distances, and no repeat runs with a second force field. Without those, it is hard to rule out that the reported shifts partly reflect simulation artifacts rather than ligand physics. This is the sort of paper that computational RNA biologists or groups working on frameshift-targeted antivirals would want to see. A reader looking for mechanistic hypotheses on topology-dependent drug action could extract useful ideas, even if the quantitative landscapes need more support. It is coherent on its own terms and engages the right literature, so it deserves a serious referee. I would send it out for review and expect the main requests to be the missing validation steps on the collective variables.

Referee Report

2 major / 1 minor

Summary. The manuscript applies molecular dynamics simulations of the SARS-CoV-2 RNA pseudoknot (threaded and unthreaded topologies) in complex with merafloxacin and two analogs, in both neutral and zwitterionic forms. Spectral map, a thermodynamics-driven machine-learning method, is used to learn collective variables directly from the trajectories. The central claims are that ligand-induced destabilization is topology-selective (S2 stem in threaded pseudoknot; S1/S3 stems in unthreaded), that the zwitterionic form uniquely imposes slow dynamics on the unthreaded pseudoknot, and that neutral versus zwitterionic forms differ qualitatively in mechanism within the same topology.

Significance. If the spectral-map-derived free-energy landscapes are robust, the work would provide concrete mechanistic insight into how small-molecule binding modulates slow conformational dynamics of a viral RNA regulatory element in a manner that depends on both pseudoknot topology and ligand protonation state. This could inform structure-based design of -1 PRF inhibitors. The data-driven extraction of CVs from unbiased MD is a methodological strength when accompanied by appropriate validation.

major comments (2)

[Methods (spectral map implementation and validation)] The central topology-selective destabilization and zwitterion-specific slow-dynamics claims rest on the spectral-map collective variables accurately isolating ligand-induced slow modes. No convergence diagnostics on the learned eigenvalues, no comparison of the resulting landscapes to those obtained with conventional collective variables (e.g., stem RMSD or base-pair distances), and no repetition with an independent RNA force field are described. These omissions leave open the possibility that the reported shifts in free-energy minima reflect force-field artifacts or incomplete sampling rather than physical ligand effects.
[Results (free-energy landscapes)] The abstract states that free-energy landscapes were constructed from the SM-projected trajectories, yet no error estimation, block-averaging, or overlap metrics between independent runs are mentioned. Without these, it is difficult to assess whether the observed differences between threaded and unthreaded topologies, or between neutral and zwitterionic ligands, exceed statistical uncertainty.

minor comments (1)

[Abstract] The abstract would benefit from explicit statement of the total simulation time, number of independent trajectories, and force-field version used, to allow immediate assessment of sampling adequacy.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed comments, which have prompted us to strengthen the validation aspects of our spectral map analysis. We address each major comment point by point below and have revised the manuscript accordingly to improve clarity and robustness.

read point-by-point responses

Referee: [Methods (spectral map implementation and validation)] The central topology-selective destabilization and zwitterion-specific slow-dynamics claims rest on the spectral-map collective variables accurately isolating ligand-induced slow modes. No convergence diagnostics on the learned eigenvalues, no comparison of the resulting landscapes to those obtained with conventional collective variables (e.g., stem RMSD or base-pair distances), and no repetition with an independent RNA force field are described. These omissions leave open the possibility that the reported shifts in free-energy minima reflect force-field artifacts or incomplete sampling rather than physical ligand effects.

Authors: We agree that explicit validation of the spectral map collective variables is essential. In the revised manuscript, we have added convergence diagnostics by monitoring the stability of the learned eigenvalues across successive blocks of each trajectory, demonstrating that the dominant modes converge within the simulation lengths. We have also included side-by-side comparisons of the free-energy landscapes projected onto the spectral map coordinates versus conventional collective variables (stem RMSD and base-pair distances), confirming that the topology-selective minima shifts are reproduced. For the RNA force field, our simulations used the standard Amber ff14SB parameters with RNA-specific modifications; we have expanded the Methods and Discussion sections to justify this choice, cite supporting benchmarks for pseudoknot systems, and explicitly note the absence of independent-force-field repetition as a limitation while arguing that the observed ligand-induced effects are consistent with known RNA-ligand interaction principles. revision: yes
Referee: [Results (free-energy landscapes)] The abstract states that free-energy landscapes were constructed from the SM-projected trajectories, yet no error estimation, block-averaging, or overlap metrics between independent runs are mentioned. Without these, it is difficult to assess whether the observed differences between threaded and unthreaded topologies, or between neutral and zwitterionic ligands, exceed statistical uncertainty.

Authors: We thank the referee for highlighting this omission. The revised Results section now incorporates error estimation on all free-energy landscapes via block averaging over non-overlapping trajectory segments, with the resulting uncertainties shown as contour shading. Where multiple independent runs were available for a given system, we report distribution overlap metrics (e.g., Jensen-Shannon divergence) between the projected ensembles. These additions confirm that the reported topology-selective destabilization and zwitterion-specific slow dynamics exceed the estimated statistical uncertainties. revision: yes

Circularity Check

0 steps flagged

Central results from independent MD trajectories via data-driven spectral map; no load-bearing reduction to fitted parameters or self-citations

full rationale

The derivation proceeds from unbiased MD trajectories of threaded/unthreaded pseudoknots with neutral/zwitterionic ligands, applies spectral map to extract CVs, then constructs free-energy landscapes and identifies topology-selective stem destabilization plus zwitterion-specific slow dynamics. No equation or claim reduces these observations to quantities defined by the method's own fitted parameters or to prior self-citations; the landscapes are direct projections of the sampled trajectories. A minor self-citation for the SM method itself is present but not load-bearing for the reported mechanistic distinctions, which remain falsifiable against the underlying simulation data.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard assumptions from molecular dynamics and the validity of the spectral map method for extracting thermodynamically relevant collective variables; no new entities are introduced and no free parameters are explicitly fitted to match the target result.

axioms (2)

domain assumption Molecular dynamics simulations with standard force fields sufficiently sample the relevant conformational space of the RNA-ligand complex on the timescales used.
Invoked implicitly when interpreting free-energy landscapes from MD trajectories.
domain assumption Spectral map correctly identifies collective variables that isolate the slowest modes from high-frequency fluctuations.
Central to the method used to generate the reported free-energy landscapes.

pith-pipeline@v0.9.0 · 5645 in / 1581 out tokens · 49030 ms · 2026-05-10T09:59:35.745472+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

14 extracted references · 14 canonical work pages

[1]

(1) Dinman, J. D. Mechanisms and Implications of Programmed Transla- tional Frameshifting.Wiley Interdiscip. Rev. RNA2012,3, 661–673, DOI: https://doi.org/10.1002/wrna.1126. (2) Falese, J. P.; Donlic, A.; Hargrove, A. E. Targeting RNA with Small Molecules: From 27 Fundamental Principles Towards the Clinic.Chem. Soc. Rev.2021,50, 2224–2243, DOI:https://doi...

work page doi:10.1002/wrna.1126 2021
[2]

N.; Heermann, R.; Kersten, C

Schwickert, M.; Hoba, S. N.; Heermann, R.; Kersten, C. Electrostatic Anchoring in RNA-Ligand Design—Dissecting the Effects of Positive Charges on Affinity, Selectivity, Binding Kinetics, and Thermodynamics.J. Med. Chem.2025,68, 8659–8678, DOI: https://doi.org/10.1021/acs.jmedchem.5c00339. (6) Finkel, Y.; Mizrahi, O.; Nachshon, A.; Weingarten-Gabbay, S.; M...

work page doi:10.1021/acs.jmedchem.5c00339 2025
[3]

(7) Jones, C

Yahalom-Ronen, Y.; Tamir, H.; Achdout, H.; Stein, D.; Israeli, O.; oth- ers The Coding Capacity of SARS-CoV-2.Nature2021,589, 125–130, DOI: https://doi.org/10.1038/s41586-020-2739-1. (7) Jones, C. P.; Ferré-D’Amaré, A. R. Crystal Structure of the Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) Frameshifting Pseudoknot.RNA2022,28, 239–249, DOI...

work page doi:10.1038/s41586-020-2739-1 2021
[4]

D.; Rangan, R.; Kladwang, W.; Li, S.; others Cryo-Em and Anti- sense Targeting of the 28-kDa Frameshift Stimulation Element From the SARS- CoV-2 RNA Genome.Nat

Pintilie, G. D.; Rangan, R.; Kladwang, W.; Li, S.; others Cryo-Em and Anti- sense Targeting of the 28-kDa Frameshift Stimulation Element From the SARS- CoV-2 RNA Genome.Nat. Struct. & Mol. Biol.2021,28, 747–754, DOI: https://doi.org/10.1038/s41594-021-00653-y. (10) Bhatt, P. R.; Scaiola, A.; Loughran, G.; Leibundgut, M.; Kratzel, A.; Meurs, R

work page doi:10.1038/s41594-021-00653-y 2021
[5]

M.; McMillan, A.; Bode, J

Dreos, R.; O’Connor, K. M.; McMillan, A.; Bode, J. W.; others Structural Basis of Ribosomal Frameshifting during Translation of the SARS-CoV-2 RNA Genome.Sci- ence2021,372, 1306–1313, DOI:https://doi.org/10.1126/science.abf3546. (11) Peterson, J. M.; Becker, S. T.; O’Leary, C. A.; Juneja, P.; Yang, Y

work page doi:10.1126/science.abf3546
[6]

Moss, W. N. Structure of the SARS-CoV-2 Frameshift Stimulatory Element With an Upstream Multibranch Loop.Biochemistry2024,63, 1287–1296, DOI: https://doi.org/10.1021/acs.biochem.3c00716. (12) Wacker, A.; Weigand, J. E.; Akabayov, S. R.; Altincekic, N.; Bains, J. K

work page doi:10.1021/acs.biochem.3c00716
[7]

Acids Res.2020,48, 12415–12435, DOI: https://doi.org/10.1093/nar/gkaa1013

Banijamali, E.; Binas, O.; Castillo-Martinez, J.; Cetiner, E.; Ceylan, B.; oth- ers Secondary Structure Determination of Conserved SARS-CoV-2 RNA Ele- ments by NMR Spectroscopy.Nucl. Acids Res.2020,48, 12415–12435, DOI: https://doi.org/10.1093/nar/gkaa1013. (13) Neupane, K.; Zhao, M.; Lyons, A.; Munshi, S.; Ileperuma, S. M.; Ritchie, D. B

work page doi:10.1093/nar/gkaa1013 2020
[8]

Q.; Narayan, A.; Woodside, M

Hoffer, N. Q.; Narayan, A.; Woodside, M. T. Structural Dynamics of Single SARS- CoV-2PseudoknotMoleculesRevealTopologicallyDistinctConformers.Nat. Commun. 2021,12, 4749, DOI:https://doi.org/10.1038/s41467-021-25085-6. (14) Pekarek, L.; Zimmer, M. M.; Gribling-Burrer, A.-S.; Buck, S.; Smyth, R

work page doi:10.1038/s41467-021-25085-6 2021
[9]

Cis-Mediated Interactions of the SARS-CoV-2 Frameshift RNA Alter 29 its Conformations and Affect Function.Nucl

Caliskan, N. Cis-Mediated Interactions of the SARS-CoV-2 Frameshift RNA Alter 29 its Conformations and Affect Function.Nucl. Acids Res.2023,51, 728–743, DOI: https://doi.org/10.1093/nar/gkac1184. (15) Omar, S. I.; Zhao, M.; Sekar, R. V.; Arbabimoghadam, S.; Tuszynski, J. A.; Wood- side, M. T. Modeling the Structure of the Frameshift-Stimulatory Pseudoknot...

work page doi:10.1093/nar/gkac1184 2023
[10]

D.; Loerch, S.; Woodside, M

Dinman, J. D.; Loerch, S.; Woodside, M. T. Identifying Inhibitors of−1 Programmed Ribosomal Frameshifting in a Broad Spectrum of Coronaviruses.Viruses2022,14, 177, DOI:https://doi.org/10.3390/v14020177. (23) Sun, Y.; Abriola, L.; Niederer, R. O.; Pedersen, S. F.; Alfajaro, M. M.; Silva Mon- teiro, V.; Wilen, C. B.; Ho, Y.-C.; Gilbert, W. V.; Surovtseva, Y...

work page doi:10.3390/v14020177 2021
[11]

Rev.2018,118, 4177– 4338, DOI:https://doi.org/10.1021/acs.chemrev.7b00427

Pinamonti, G.; Poblete, S.; Jurecka, P.; others RNA Structural Dynamics as Captured by Molecular Simulations: A Comprehensive Overview.Chem. Rev.2018,118, 4177– 4338, DOI:https://doi.org/10.1021/acs.chemrev.7b00427. 32 (34) Mlynsky, V.; Bussi, G. Exploring RNA Structure and Dynamics Through En- hanced Sampling Simulations.Curr. Opin. Struct. Biol.2018,49,...

work page doi:10.1021/acs.chemrev.7b00427 2018
[12]

P.; Le Grice, S

Seshadri, S.; Liang, X.; Jones, C. P.; Le Grice, S. F.; Ferre-D’Amare, A. R.; others Discovery of Small Molecules Targeting the Frameshifting Element RNA in SARS-CoV-2 Viral Genome.ACS Med. Chem. Lett.2023,14, 757–765, DOI: https://doi.org/10.1021/acsmedchemlett.3c00051. (49) Trott, O.; Olson, A. J. AutoDock Vina: Improving the Speed and Accuracy of Docki...

work page doi:10.1021/acsmedchemlett.3c00051 2023
[13]

Darian, E.; Guvench, O.; Lopes, P.; Vorobyov, I.; others CHARMM General Force Field: A Force Field for Drug-like Molecules Compatible with the CHARMM All- Atom Additive Biological Force Fields.J. Comput. Chem.2010,31, 671–690, DOI: https://doi.org/10.1002/jcc.21367. (54) Bussi, G.; Donadio, D.; Parrinello, M. Canonical Sampling Through Velocity Rescaling....

work page doi:10.1002/jcc.21367 2010
[14]

Neural Inf

Lin, Z.; Gimelshein, N.; Antiga, L.; others PyTorch: An Imperative Style, High- Performance Deep Learning Library .Adv. Neural Inf. Process. Syst. 33, 8026–8037. (59) Maragliano, L.; Fischer, A.; Vanden-Eijnden, E.; Ciccotti, G. String Method in Col- lective Variables: Minimum Free Energy Paths and Isocommittor Surfaces.J. Chem. Phys.2006,125, 024106, DOI...

work page doi:10.1063/1.2212942 2006