Pith. sign in

REVIEW 3 major objections 4 minor 30 references

VTX: Real-time high-performance molecular structure and dynamics visualization software

T0 review · 3 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read VTX is an open-source molecular visualization program claiming that a 101,753,154-bead Martini whole-cell model can be loaded, selected, and freely navigated in real time on a consumer laptop with an NVIDIA RTX2080-class GPU.

desk verdict Useful open-source molecular viewer with an impressive whole-cell demo, but the 'real-time' claim is backed by self-graded qualitative labels, not measured frame times. read the letter →

arxiv 2501.12750 v2 pith:ZCK4US2K submitted 2025-01-22 cs.GR physics.chem-phq-bio.BM

Maxime Maria , Simon Guionnière , Nicolas Dacquay , Cyprien Plateau-Holleville , Valentin Guillaume , Vincent Larroque , Jean Lardé , Yassine Naimi
show 5 more authors
Jean-Philip Piquemal Guillaume Levieux Nathalie Lagarde Stéphane Mérillou Matthieu Montes
This is my paper · ORCID
classification cs.GRphysics.chem-phq-bio.BM
keywords molecularvisualizationgraphicsreal-timerenderingmeshlessrepresentationsdynamicswhole-cellmodelinteractivenavigationstructuralbiology
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

VTX is a molecular visualization program whose central claim is that massive molecular systems, up to a 101,753,154-bead Martini whole-cell model, can be loaded, selected, and navigated in real time on a consumer laptop. The paper attributes this to meshless representations: atoms, bonds, and surfaces are drawn as implicit primitives on quads, ray-cast in the shader, so no triangle mesh has to be stored or streamed. On benchmark systems ranging from 107,640 atoms to 101 million beads, VTX reports stable loading and the highest interactivity grade on every system, while ChimeraX crashes, PyMOL freezes, and VMD stalls on the largest one. A sympathetic reader would take the contribution to be a working, open-source answer to the gap between simulation data size and interactive molecular graphics, useful for research, teaching, and illustration.

What carries the argument

The key machinery is the meshless molecular graphics engine: instead of building triangle meshes, VTX stores each representation (sticks, ball-and-stick, Van der Waals, solvent-accessible surface) as implicit geometric primitives. Each primitive is drawn as a simple quad and the fragment shader ray-casts the implicit equation to produce a pixel-perfect shape, which drastically lowers memory consumption and bandwidth. A deferred shading pipeline computes lighting and screen-space effects such as ambient occlusion in real time; cartoon representations are generated on the fly with adaptive level-of-detail using tessellation shaders, and solvent-excluded surfaces are generated by a discrete method and marching cubes. This combination is what the paper argues lets a laptop GPU hold and redraw scenes with tens of millions of primitives.

What would settle it

Record per-frame GPU timings while a scripted sequence orbits, zooms, selects, and toggles representations in the 101,753,154-bead whole-cell model on the same RTX2080 Super laptop; if sustained frame times exceed roughly 33 ms (30 fps) during those operations, or load-time and memory measurements show triangle-mesh-equivalent scaling, the real-time scalability claim is falsified.

Watch

Extended reading notes

Core claim

The paper's central claim is that a meshless rendering engine can make whole-cell-scale molecular models interactively explorable on hardware available in a normal lab. In the authors' benchmark, VTX loads all four test systems, from a 107,640-atom pilus machine to a 101,753,154-bead Martini minimal whole-cell model, on a laptop with an RTX2080 Super; on the largest system it reports full interactive fluidity for selection, representation changes, and manipulation. The three comparison tools fail at that scale: ChimeraX crashes, PyMOL freezes during loading, and VMD, though it loads the system, freezes when the user selects or manipulates it. The paper therefore presents VTX as a practical bridge between modern simulation data sizes and the interactive tools structural biologists actually use.

Load-bearing premise

The paper's performance case rests on a single laptop and on the authors' qualitative fluidity grades (+, ++, +++) in lieu of frame-rate measurements, error bars, or an independent protocol, while assuming the four benchmark systems represent massive molecular systems generally.

Editorial extensions

If this is right

  • A whole-cell simulation with 101,753,154 coarse-grained beads can be interactively explored with free-fly navigation, precise selection, and live representation changes on a single laptop rather than a visualization cluster.
  • Million-atom single structures and multi-million-atom assemblies can be loaded and manipulated without precomputed meshes, so memory and bandwidth scale with scene content rather than triangle count.
  • Researchers using Martini-style coarse-grained simulations can inspect whole-cell trajectories directly, a step toward analyzing the output of next-generation cell-scale molecular dynamics runs.
  • Because VTX reads common file formats through the Chemfiles library and is freely available for non-commercial use, the same rendering pipeline can be dropped into existing structural biology and teaching workflows.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • An implication the authors leave implicit is that the meshless approach could be combined with out-of-core streaming to handle systems larger than GPU memory, since quad generation is cheap; the reported whole-cell test fits in available memory, so the limit is not yet probed.
  • If the qualitative fluidity grades were replaced with frame-time traces, the scalability claim could become a reproducible benchmark across hardware; many users will care whether VTX stays interactive on integrated or older GPUs as well as on an RTX2080 Super.
  • The free-fly camera and keyboard-first controls already resemble first-person game navigation, so an immersive or virtual-reality front-end would be a natural extension.
  • The success on the Martini whole-cell model suggests VTX could serve as an interactive debugging and visualization environment for coarse-grained simulations of large cellular machines, not only for atomic-resolution structures.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper presents VTX, an open-source molecular visualization software aimed at real-time rendering and interactive manipulation of very large molecular systems and MD trajectories. The engine uses meshless (impostor-based) representations with deferred shading, a modular Qt GUI, free-fly and trackball cameras, and the chemfiles library for format compatibility. The authors benchmark VTX version 0.4.4 against VMD, PyMOL, and ChimeraX on four systems up to a 101,753,154-bead Martini minimal whole-cell model, reporting system stability and qualitative interactive-fluidity grades on a single Dell Alienware laptop with an RTX 2080 Super.

Significance. If the central performance claim is substantiated, VTX would address a genuine and growing need: interactive visualization of whole-cell-scale molecular dynamics data on commodity hardware. The paper is well-written, the software is open source, it builds on a solid base (chemfiles, modern OpenGL), and the authors provide a video demonstration and a public benchmark deposit for the smaller test systems. The main weakness is that the headline scalability result rests entirely on self-assigned qualitative fluidity grades with no quantitative frame-time data and no publicly accessible copy of the 100M-bead benchmark system, which prevents independent verification. Because the contribution is precisely the claimed real-time capability, this evidence gap is load-bearing.

major comments (3)
  1. [Performance evaluation / Table 1] The core claim of real-time rendering of the 101,753,154-bead whole-cell model is supported only by categorical fluidity grades ('+', '++', '+++') assigned by the authors. No FPS, frame-time, or frame-time-variance numbers are reported, no measurement protocol is given, and the text does not state how many runs were performed, which specific manipulation tasks were timed, or how the grades were mapped to user experience. Because the entire contribution is the claimed scalability, this is load-bearing evidence. Please report quantitative metrics (e.g., average and 1st/99th percentile frame times during camera rotation, selection, and representation changes) and describe a repeatable protocol, including settings used for all compared tools.
  2. [Data and software availability] The manuscript states that 'All files composing the benchmarking dataset, except the Martini minimal whole cell model, are available at https://doi.org/10.5281/zenodo.14962673' and that the .gro files for the whole-cell model are 'available upon request at the Marrink lab.' Since the whole-cell model is the key test case for the headline result, the benchmark cannot be independently reproduced or verified. Please make the 100M-bead system publicly available in a persistent repository, or, if redistribution is restricted, provide a representative equivalent system of the same size and complexity that can be used to reproduce the reported fluidity results.
  3. [Molecular Graphics Engine] The rendering description says that for each primitive a simple quad is rasterized, followed by ray-casting of the implicit surface. With 100M beads, this implies processing on the order of 100M quads per frame unless there is aggressive culling, level-of-detail, instancing, or another aggregation strategy. The manuscript does not describe any such mechanism, so the claimed interactivity is non-obvious and currently unexplained. Please specify the rendering stages (e.g., frustum culling, LOD selection, instancing, or point-based pre-pass) and report measured draw-call counts, vertex/fragment workload, and GPU memory usage for the 100M system.
minor comments (4)
  1. [Use case] The text refers to a 'NVIDIA RTX2080m' while Table 1 and the Performance evaluation section specify an 'RTX2080 Super'; please use one consistent GPU designation.
  2. [Table 1 caption] The caption says 'Interactive stability measures...' but the column is labeled 'Interactive Fluidity'; align the caption terminology with the column header.
  3. [Data and software availability] The paper states that VTX is 'open source' but also that it is 'free for non commercial use' under a 'VTX consortium license'; if the license restricts commercial use, the term 'open source' may be misleading and should be clarified or replaced with 'source-available.'
  4. [Availability] The supplementary video is mentioned in the abstract and text, but no direct link or accession is provided in the manuscript PDF; please add the URL or DOI in the Data and software availability section.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper reports an empirical software benchmark; the central scalability claim rests on self-assessed qualitative fluidity grades, which is an evidence-quality concern, not a circular derivation.

full rationale

VTX is a software description paper. Its central claim is that VTX allows real-time rendering and manipulation of massive molecular systems, including a 101,753,154-bead whole-cell model on a consumer laptop. That claim is supported by a comparative performance table with qualitative fluidity grades assigned by the authors on a single machine. This is a legitimate empirical observation about the software, not a derivation that reduces to its own inputs by construction. There is no fitted parameter renamed as a prediction, no uniqueness theorem imported from the authors' prior work, and no ansatz smuggled in via a self-citation. The references to the authors' own prior work (UDock2 [20] and the SES paper [29]) are contextual or forward-looking and are not load-bearing for the real-time scalability result. The absence of frame-time measurements and the self-graded nature of Table 1 raise reproducibility and evidence-quality concerns, but those are not circularity under the specified patterns. The benchmark systems come from external sources and the compared tools are independent. Therefore, no specific circular step can be quoted and exhibited, and the honest finding is no significant circularity (score 0).

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The central performance claim rests on three unverified domain assumptions about the benchmark: that the chosen test systems represent large molecular data generally, that the subjective fluidity grades are meaningful, and that the single benchmark machine represents consumer hardware. No new physical or mathematical entities are introduced.

assumptions (3)
  • domain assumption The four benchmark systems are representative of massive molecular systems generally.
    Used to generalize from one whole-cell model and three large structures to a broad claim about very large molecular systems. Enter in the Performance evaluation section.
  • domain assumption The subjective fluidity grades (+, ++, +++) are reliable, consistent measures of interactive responsiveness.
    No inter-rater or objective validation is provided. Enter in Table 1.
  • domain assumption The Dell Alienware 15r with an RTX 2080 Super is a typical consumer laptop.
    The performance claims rely on this specific hardware as the benchmark platform. Enter in the Performance evaluation section.

how reviews work

0 comments
Cite this review

Pith. "Pith review of VTX: Real-time high-performance molecular structure and dynamics visualization software." pith.science (2026). https://pith.science/paper/ZCK4US2K

@misc{pith2026250112750,
  author       = {Pith},
  title        = {Pith review of: VTX: Real-time high-performance molecular structure and dynamics visualization software},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZCK4US2K}},
  note         = {Machine review of arXiv:2501.12750}
}
read the original abstract

Summary: VTX is a molecular visualization software capable to handle most molecular structures and dynamics trajectories file formats. It features a real-time high-performance molecular graphics engine, based on modern OpenGL, optimized for the visualization of massive molecular systems and molecular dynamics trajectories. VTX includes multiple interactive camera and user interaction features, notably free-fly navigation and a fully modular graphical user interface designed for increased usability. It allows the production of high-resolution images for presentations and posters with custom background. VTX design is focused on performance and usability for research, teaching and educative purposes. Availability and implementation: VTX is open source and free for non commercial use. Builds for Windows and Ubuntu Linux are available at http://vtx.drugdesign.fr. The source code is available at https://github.com/VTX-Molecular-Visualization . Supplementary Information: A video displaying free-fly navigation in a whole-cell model is available

Figures

Figures reproduced from arXiv: 2501.12750 by the authors.

Figure 1
Figure 1. A. Illustration of the VTX GUI organized with different dockable panels. The Scene Tree panel contains and allows the selection of all the objects present in the scene, such as molecules, labels and viewpoints. The Sequence panel displays and allows the selection of the sequence of each biomolecule , while the Visualization Panel allows the selection, observation, and manipulation of their 3D structure. Frequently u… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 29 canonical work pages

  1. [1]

    A. J. Olson. Perspectives on structural molecular biology visualization: From past to present.Journal of Molecular Biology, 430:3997–4012, 2018

    work page 2018

  2. [2]

    Atomic-resolution protein structure determination by cryo-em

    John Jumper, Richard Evans, et al. Atomic-resolution protein structure determination by cryo-em. Nature, 596:583–589, 2021

    work page 2021

  3. [3]

    K.M. Yip, N. Fischer, et al. Highly accurate protein structure prediction with alphafold. Nature, 587:157–161, 2020

    work page 2020

  4. [4]

    AlphaFold Protein Structure Database: massively expanding the structural coverage of protein-sequence space with high-accuracy models

    Mihaly Varadi, Stephen Anyango, et al. AlphaFold Protein Structure Database: massively expanding the structural coverage of protein-sequence space with high-accuracy models. Nucleic Acids Research, 50(D1):D439–D444, 11 2021

    work page 2021

  5. [5]

    The need to implement FAIR principles in biomolecular simulations, August

    Rommie Amaro, Johan Åqvist, et al. The need to implement FAIR principles in biomolecular simulations, August

    work page

  6. [6]

    MD-DATA: the legacy of the ABC Consortium

    Adam Hospital and Modesto Orozco. MD-DATA: the legacy of the ABC Consortium. Biophysical Reviews, 16(3):269–271, June 2024

    work page 2024

  7. [7]

    Scaling molecular dynamics beyond 100,000 processor cores for large-scale biophysical simulations

    Jaewoon Jung, Wataru Nishima, et al. Scaling molecular dynamics beyond 100,000 processor cores for large-scale biophysical simulations. Journal of Computational Chemistry, 40(21):1919–1930, 2019

    work page 1919

  8. [8]

    Stevens, Fabian Grünewald, et al

    Jan A. Stevens, Fabian Grünewald, et al. Molecular dynamics simulation of an entire cell. Frontiers in Chemistry, 11, 2023

    work page 2023

Show all 30 references
  1. [9]

    The structure and physical properties of a packaged bacteriophage particle

    Kush Coshic, Christopher Maffeo, David Winogradoff, and Aleksei Aksimentiev. The structure and physical properties of a packaged bacteriophage particle. Nature, 627:905–914, 03/2024 2024

    2024

  2. [10]

    LLC Schrödinger and Warren DeLano. Pymol. 2020

    2020

  3. [11]

    Humphrey, A

    W. Humphrey, A. Dalke, and K. Schulten. Vmd: visual molecular dynamics. Journal of molecular graphics, 14(1):33–38, 1996

    1996

  4. [12]

    E. F. Pettersen, T. D. Goddard, et al. Ucsf chimerax: Structure visualization for researchers, educators, and developers. Protein Science, 30(1):70–82, 2021

    2021

  5. [13]

    Chemfiles, a modern library for chemistry file reading and writing, 2021

    Guillaume Fraux. Chemfiles, a modern library for chemistry file reading and writing, 2021

    2021

  6. [14]

    Bagur, Nithin Shivashankar, and Vijay Natarajan

    Pranav D. Bagur, Nithin Shivashankar, and Vijay Natarajan. Improved quadric surface impostors for large bio-molecular visualization. In Proceedings of the Eighth Indian Conference on Computer Vision, Graphics and Image Processing, ICVGIP ’12, New York, NY , USA, 2012. Associat...

    2012

  7. [15]

    Hermosilla, V

    P. Hermosilla, V . Guallar, A. Vinacua, and P. P. Vázquez.Instant visualization of secondary structures of molecular models. VCBM ’15. Eurographics Association, 2015

    2015

  8. [16]

    Ligand excluded surface: A new type of molecular surface

    Norbert Lindow, Daniel Baum, and Hans-Christian Hege. Ligand excluded surface: A new type of molecular surface. IEEE Transactions on Visualization and Computer Graphics, 20(12):2486–2495, 2014

    2014

  9. [17]

    Lorensen and Harvey E

    William E. Lorensen and Harvey E. Cline. Marching cubes: A high resolution 3d surface construction algorithm. SIGGRAPH Comput. Graph., 21(4):163–169, August 1987

    1987

  10. [18]

    Finding next gen: Cryengine 2

    Martin Mittring. Finding next gen: Cryengine 2. In ACM SIGGRAPH 2007 Courses, SIGGRAPH ’07, page 97–121, 2007

    2007

  11. [19]

    Udock, the interactive docking entertainment system

    Guillaume Levieux, Guillaume Tiger, et al. Udock, the interactive docking entertainment system. Faraday Discuss., 169:425–441, 2014

    2014

  12. [20]

    Udock2: interactive real-time multi-body pro- tein–protein docking software

    Cyprien Plateau-Holleville, Simon Guionnière, Benjamin Boyer, Brian Jiménez-Garcia, Guillaume Levieux, Stéphane Mérillou, Maxime Maria, and Matthieu Montes. Udock2: interactive real-time multi-body pro- tein–protein docking software. Bioinformatics, 39(10):btad609, 10 2023

    2023

  13. [21]

    Le Muzic, L

    M. Le Muzic, L. Autin, J. Parulek, and I. Viola. cellview: a tool for illustrative and multi-scale rendering of large biomolecular datasets. Eurographics Workshop Vis Comput Biomed., pages 61–70, 2015

    2015

  14. [22]

    Berman, John Westbrook, et al

    Helen M. Berman, John Westbrook, et al. The Protein Data Bank. Nucleic Acids Research, 28(1):235–242, January 2000

    2000

  15. [23]

    Qt : cross-platform application and ui framework

    Nokia Corp. Qt : cross-platform application and ui framework. 2012

    2012

  16. [24]

    Rettberg, Anke Treuner-Lange, Janet Iwasa, Lotte Søgaard-Andersen, and Grant J

    Yi-Wei Chang, Lee A. Rettberg, Anke Treuner-Lange, Janet Iwasa, Lotte Søgaard-Andersen, and Grant J. Jensen. Architecture of the type iva pilus machine. Science, 351(6278):aad2001, 2016. 6 VTX: Real-time high-performance molecular structure and dynamics visualization software

    2016

  17. [25]

    Structure and genome ejection mechanism of <i>staphylococcus aureus</i> phage p68

    Dominik Hrebík, Dana Štveráková, Karel Škubník, Tibor Füzik, Roman Pant˚ uˇcek, and Pavel Plevka. Structure and genome ejection mechanism of <i>staphylococcus aureus</i> phage p68. Science Advances, 5(10):eaaw7414, 2019

    2019

  18. [26]

    Axonemal structures reveal mechanoregulatory and disease mechanisms

    Travis Walton, Miao Gui, Simona Velkova, and Alan Brown. Axonemal structures reveal mechanoregulatory and disease mechanisms. Nature, 2023

    2023

  19. [27]

    Bayfield, Andrey N

    Oliver W. Bayfield, Andrey N. Shkoporov, Natalya Yutin, and Alfred A. Antson. Structural atlas of a human gut crassvirus. Nature, 2023

    2023

  20. [28]

    Protein structure alignment by incremental combinatorial extension (ce) of the optimal path

    Ilya N Shindyalov and Philip E Bourne. Protein structure alignment by incremental combinatorial extension (ce) of the optimal path. Protein engineering, 11(9):739–747, 1998

    1998

  21. [29]

    Efficient gpu computation of large protein solvent-excluded surface

    Cyprien Plateau—Holleville, Maxime Maria, Stéphane Mérillou, and Matthieu Montes. Efficient gpu computation of large protein solvent-excluded surface. IEEE Transactions on Visualization and Computer Graphics, pages 1–12, 2024. 7

    2024

  22. [2024]

    arXiv:2407.16584 [q-bio]

    arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.