Pith. sign in

REVIEW 3 major objections 6 minor 68 references

Machine Learning-driven Multiscale MD Workflows: The Mini-MuMMI Experience

T0 review · 3 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Mini-MuMMI shows that a compact ML-driven workflow can bridge two protein states in a 36-hour campaign.

desk verdict A useful open-source workflow packaging with real performance numbers; the 'bridging the gap' claim is a 1D histogram artifact, not a demonstrated conformational path. read the letter →

arxiv 2507.07352 v1 pith:I5FKPQG4 submitted 2025-07-10 cs.DC cs.LG

Loïc Pottier , Konstantia Georgouli , Timothy S. Carpenter , Fikret Aydin , Jeremy O. B. Tempkin , Dwight V. Nissley , Frederick H. Streitz , Thomas R. W. Scogland
show 3 more authors
Peer-Timo Bremer Felice C. Lightstone Helgi I. Ingólfsson
This is my paper
classification cs.DCcs.LG
keywords multiscaleworkflowmoleculardynamicsmachinelearningautoencoderlatentspacecoarse-grainedsimulationRAS-RAFHPCmanagementmini-MuMMI
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

This paper introduces mini-MuMMI, a smaller, open-source version of the MuMMI multiscale workflow that coordinates machine-learning-guided coarse-grained molecular dynamics simulations. It aims to establish that the infrastructure can be packaged to run on modest HPC systems while remaining fully functional, and that such a workflow can generate plausible intermediate protein conformations between two endpoint states. On 64 nodes of Frontier, mini-MuMMI orchestrated 1,983 generated structures and aggregated more than 1.1 milliseconds of coarse-grained simulation in 36 hours. The authors present this as evidence that ML-driven multiscale ensemble workflows are accessible enough for other groups to adopt, adapt, and build upon.

What carries the argument

The load-bearing object is the autoencoder and its low-dimensional latent space, which compresses the correlated motions of the two endpoint ensembles A and B. The sampler interpolates points in this latent space, the decoder turns each latent point into a coarse-grained protein structure, and the validation module accepts only structures that pass a short vacuum energy-minimization threshold with GROMACS. A workflow manager built on Maestro and Flux schedules createsims and CGAnalysis jobs, with RabbitMQ used for inter-job messaging and feedback files steering the sampler between exploration and exploitation.

What would settle it

Take the generated intermediate structures from the bridge region and run longer, unbiased coarse-grained simulations starting from them; if the CRD distance distribution collapses back to the two endpoint peaks rather than remaining in the intermediate region, then the structures are not stable intermediates. A stronger test is to re-cross a sample of generated structures to all-atom resolution and run short atomistic simulations to see whether they relax to physically plausible RAS-RAF-membrane complexes.

Watch

Extended reading notes

Core claim

The central claim is that mini-MuMMI, a curated version of the MuMMI workflow, is a completely functional, deployable template for ML-driven multiscale MD, and that a small campaign is enough to bridge two conformational states. The authors trained a 32-dimensional autoencoder latent space on the two end-state ensembles of KRAS-RBDCRD, sampled points between them, decoded those points into coarse-grained structures, validated them with short vacuum energy minimization, and ran GROMACS simulations with a Martini 3 membrane. The campaign generated 1,983 valid new structures whose CRD-membrane distances fill the gap between ensemble A (CRD distance at least 4.2 nm) and ensemble B (CRD distance at most 3.2 nm). Aggregating 1,109.4 microseconds of coarse-grained simulation, with 1,632 of 1,983 simulations reaching the 600 ns cap, is put forward as demonstration that the workflow can scale enough to support this style of pathway exploration.

Load-bearing premise

That the autoencoder latent space trained only on the two endpoint ensembles encodes a physically meaningful interpolation between them, so that decoded structures are plausible starting points rather than artificial chimeras.

Editorial extensions

If this is right

  • If mini-MuMMI works as claimed, a group with a small HPC allocation can run an ML-driven multiscale campaign without building the infrastructure from scratch.
  • The performance numbers (createsims about 584 seconds, CGAnalysis about 2097 ns/day on one Frontier GCD, validation about 3.5 seconds per structure) give workflow and scheduler researchers realistic job statistics to test resource-management algorithms.
  • Because mini-MuMMI removes the macro-model and uses a smaller system, the same orchestration pattern can be retargeted to other proteins by swapping the pre-trained autoencoder and the GROMACS setup.
  • The demonstration implies that latent-space interpolation followed by CG simulation can produce structures that populate intermediate values of a collective variable such as CRD distance, not just the endpoints.

Reading between the lines

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

  • A testable extension would be to re-cross the generated intermediate structures back to atomistic resolution and check whether they remain stable, which would test whether the latent space encodes a physically meaningful pathway rather than an artificial interpolation.
  • The workflow's separation of application and coordination layers suggests the same orchestration machinery could drive other ensemble-based ML-simulation loops, such as weighted ensemble sampling or active learning for materials, though the paper only demonstrates MD.
  • The CRD-distance filling alone does not prove a transition path; an independent check, such as computing committor probabilities or comparing against experimentally observed RAS-RAF conformations, would be needed to distinguish sampling the gap from sampling a genuine connecting pathway.
  • If the energy-minimization validation threshold is too permissive or too strict, the acceptance rate of generated structures becomes a controllable bias; reporting the acceptance rate across the campaign would let other groups calibrate their own validation filters.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper introduces mini-MuMMI, a curated and open-source version of the MuMMI multiscale workflow for machine-learning-driven molecular dynamics. It describes the two-layer architecture (application and coordination), the structure generator (sampler, generator, validation module), the simulation setup (createsims, CGAnalysis), and the orchestration stack (Maestro, Flux, RabbitMQ). A demonstration campaign on 64 Frontier nodes ran for 36 hours, orchestrated 1,983 coarse-grained simulations, aggregated about 1.1 ms of simulation time, and produced structures spanning intermediate CRD–membrane distances between two end-state ensembles. The paper also reports performance statistics (Table 1) and derives lessons for HPC support of multiscale workflows.

Significance. The workflow itself is a valuable contribution: it is open-source, reproducible via Spack, and operates at a scale accessible to small HPC clusters. The performance data and architecture description will be useful to the workflow and HPC scheduling communities. However, the scientific demonstration of 'bridging' two conformational states is not yet established: it rests on a 1D order-parameter histogram of the same coordinate used to define the states, with only vacuum energy minimization as a physical filter. The biological claims therefore need either independent validation or a more modest interpretation.

major comments (3)
  1. [Section 4.1, Figure 5C] The 'bridging the two states' claim is supported only by a histogram of the CRD–membrane distance, which is the same order parameter used to define ensembles A and B. Because the autoencoder is trained on A and B and the sampler interpolates in latent space between their encodings, intermediate CRD values are expected by construction. This does not demonstrate a conformational pathway. The only filter is vacuum energy minimization (Section 3.1), which checks for gross clashes, not physical plausibility of a transition path. Please either add independent validation (e.g., atomistic re-crossing, free-energy analysis, or comparison with known RAS-RAF structures) or explicitly limit the claim to 'generated structures with intermediate CRD distances.'
  2. [Sections 2.2 and 3.1] The ML sampling procedure is described only as 'interpolating between points from A and B after being encoded into the LS representation,' without specifying the autoencoder architecture, training data size, latent dimension choices, or the exploration/exploitation feedback mechanism. Since the bridge claim depends on the latent space being physically meaningful, this omission prevents readers from assessing or reproducing the core ML step. Please provide these details or a specific pointer to a reproducible configuration in the released repository.
  3. [Section 4.1 footnote] The paper states that papers analyzing data from this campaign have not yet been publicly released. This means the biological conclusions of the demonstration are based on unpublished analyses. To make the demonstration self-contained, please either include the relevant analysis in the manuscript or present the campaign explicitly as a workflow stress-test rather than a biological discovery.
minor comments (6)
  1. [Section 2.3] The statement that mini-MuMMI is 'not different from the full version of MuMMI from a workflow perspective – same jobs, identical data flow' is overstated because mini-MuMMI lacks the macro model and UCG mode; please reword to 'conceptually identical workflow.'
  2. [Table 1] The medium-system CGAnalysis numbers use ddcMD rather than GROMACS; consider adding a column for the MD engine so the comparison is clearer.
  3. [Section 4.1] The text uses 'about 1.1 milliseconds' and '1109.4 microseconds' interchangeably; using milliseconds consistently would improve readability.
  4. [Figure 5C] Please clarify in the caption whether the 'Generated structures' panel shows the 1,983 validated structures or a subset.
  5. [References] Reference [13] lacks a publication venue and year; please complete the bibliographic entry.
  6. [Abstract] The abstract states 'spanning from millisecond to nanosecond'; the conventional ordering would be 'nanosecond to millisecond.'

Circularity Check

2 steps flagged · score 6.0 of 10

The 'bridging the two states' result is partly by construction: the autoencoder is trained on endpoint ensembles defined by CRD distance, and the success metric is the same CRD-distance histogram, so intermediate values are expected from interpolation rather than demonstrated as a physical path.

  1. fitted input called prediction [Section 4.1, 'Simulation Results', Figure 5C; cf. Section 3.1]
    "As shown in the top frame of 5C, the CRD membrane distance, defined as the distance of the CRD domain from the plasma membrane (PM) center, is used to distinguish the structures in the two ensembles [22]. By training an autoencoder using these simulation ensembles, we created a 32-dimensional latent space. This latent space was then used by SG (see Section 3.1) to generate new protein structures between A and B. As a result, a total of 1,983 valid new structures were generated, bridging the two states."

    Ensembles A and B are defined solely by thresholds on the CRD membrane distance (A: at least 4.2 nm; B: at most 3.2 nm, per the Figure 5C caption), and the autoencoder is trained only on frames from those two endpoint ensembles (Section 2.2). The reported evidence of bridging is the CRD-distance distribution of generated structures and simulation frames (Figure 5C), which is the same order parameter used to define the training inputs. Section 3.1 states that the sampler generates points by interpolating between points from A and B after encoding them into the latent space. For a smoothly trained decoder, such interpolation will populate intermediate CRD values almost automatically, whether or not the decoded structures lie on a real conformational path.

  2. self definitional [Section 3.1, Application Layer, structure generator]
    "The sampler is responsible for selecting novel points to sample between the A and B states by interpolating between points from A and B after being encoded into the LS representation."

    This sampling rule makes the intermediate result a direct product of the input data: the latent space is trained on ensembles A and B, and the generated points are interpolations between those endpoints in latent space. Since the metric used to define A and B is the same CRD distance used to display the result, intermediate values in Figure 5C are expected by construction rather than evidence of a discovered transition path.

full rationale

The paper's main engineering contribution, a functional and deployable mini-MuMMI workflow with reproducible Spack packaging, Maestro/Flux orchestration, and a demonstrated 36-hour Frontier campaign producing 1,983 CG simulations and about 1.1 milliseconds of aggregate sampling, is independently evidenced and does not reduce to its inputs. The circularity is confined to the scientific demonstration that the workflow 'bridges the gap between two conformational states.' That demonstration uses the same CRD membrane distance that defines states A and B as both the training label and the success metric, and the sampler explicitly interpolates between encoded A and B points; intermediate CRD values are therefore built into the sampling construction rather than discovered as a physical pathway. No uniqueness theorem is invoked, and the self-citations to the group's earlier MuMMI papers ([13, 14, 15, 22, 25]) are normal prior-work references rather than load-bearing substitutions for evidence. The workflow-capability claims should be evaluated on their own, but the biological bridge claim needs an independent order parameter or pathway validation before it can be accepted as more than a by-construction histogram fill.

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

The paper introduces no new physical entities. The central workflow claim rests on domain assumptions about the validity of the ML latent space, the coarse-grained force field, and the energy-minimization validation criterion. The free parameters listed are configuration choices that shape the demonstration but are not fitted to an external benchmark.

free parameters (3)
  • CRD distance thresholds for ensembles A and B = A: CRD distance >= 4.2 nm; B: <= 3.2 nm
    Hand-chosen cutoffs define the two endpoint states and determine which generated structures are counted as bridging; from Section 4.1.
  • Latent space dimension = 32
    Dimension of the autoencoder latent space chosen by the authors and not justified in this paper; details deferred to [22]; from Section 4.1.
  • CG simulation length cap = 600 ns
    Mini-MuMMI is configured to stop simulations at 600 ns; this cap shapes the distribution in Figure 5B and the aggregated simulation time; from Section 4.1 and Figure 5B.
assumptions (3)
  • domain assumption Autoencoder latent space learned from A and B is a meaningful coordinate system for interpolating conformational states.
    Invoked in Section 2.1: MuMMI samples the latent space to find points on a path between A and B; the generated structures and the bridge claim depend on this assumption.
  • domain assumption Coarse-grained Martini 3 simulations with an 8-component plasma membrane capture the RAS-RAF membrane interactions relevant to the problem.
    Used in Section 3.1 (createsims), where the pre-configured system is built from Martini 3 and insane; all workflow demonstrations use these simulations.
  • ad hoc to paper A short vacuum energy minimization with an energy threshold is sufficient to validate ML-generated structures as simulation starting points.
    Defined in Section 3.1 validation module: structures that minimize below threshold are declared valid, but the threshold value is not given and no further physical validation is performed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Machine Learning-driven Multiscale MD Workflows: The Mini-MuMMI Experience." pith.science (2026). https://pith.science/paper/I5FKPQG4

@misc{pith2026250707352,
  author       = {Pith},
  title        = {Pith review of: Machine Learning-driven Multiscale MD Workflows: The Mini-MuMMI Experience},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/I5FKPQG4}},
  note         = {Machine review of arXiv:2507.07352}
}
read the original abstract

Computational models have become one of the prevalent methods to model complex phenomena. To accurately model complex interactions, such as detailed biomolecular interactions, scientists often rely on multiscale models comprised of several internal models operating at difference scales, ranging from microscopic to macroscopic length and time scales. Bridging the gap between different time and length scales has historically been challenging but the advent of newer machine learning (ML) approaches has shown promise for tackling that task. Multiscale models require massive amounts of computational power and a powerful workflow management system. Orchestrating ML-driven multiscale studies on parallel systems with thousands of nodes is challenging, the workflow must schedule, allocate and control thousands of simulations operating at different scales. Here, we discuss the massively parallel Multiscale Machine-Learned Modeling Infrastructure (MuMMI), a multiscale workflow management infrastructure, that can orchestrate thousands of molecular dynamics (MD) simulations operating at different timescales, spanning from millisecond to nanosecond. More specifically, we introduce a novel version of MuMMI called "mini-MuMMI". Mini-MuMMI is a curated version of MuMMI designed to run on modest HPC systems or even laptops whereas MuMMI requires larger HPC systems. We demonstrate mini-MuMMI utility by exploring RAS-RAF membrane interactions and discuss the different challenges behind the generalization of multiscale workflows and how mini-MuMMI can be leveraged to target a broader range of applications outside of MD and RAS-RAF interactions.

Figures

Figures reproduced from arXiv: 2507.07352 by the authors.

Figure 1
Figure 1. The MuMMI ML-driven multiscale approach. MuMMI aims to find the sequences of conformational states (i.e., paths) connecting two states A and B. MuMMI samples the latent space created by the underlying ML model to find interesting simulations to explore. MuMMI ML model is trained on frames produced by simulations modeling A and B manually set up by domain scientists. The workflow is responsible to coordinate the exec… view at source ↗
Figure 2
Figure 2. A typical example of a MuMMI campaign. The green part denotes what is managed by the MuMMI infrastructure, the orange part is what domain scientists perform manually, outside of the workflow. The initial simulations for each state are defined and executed manually, as well as the training of the ML model. Once MuMMI has a trained model, it can start exploring the conformational space between states. When all previou… view at source ↗
Figure 3
Figure 3. The mini-MuMMI workflow is comprised of two main components: the structure generator (SG) in pink, and the CG simulations (green). These components operate in sequence: starting from the SG producing ML-generated structures to the CG simulations module simulating them. The sampler within the SG consumes feedback files generated by the simulations to iteratively improve its sampling procedure. The workflow manager is… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Mini-MuMMI dataflow. The sequence is the following: (1) the WM starts by requesting new structures to explore to Structure Generator (SG), (2) the SG replies with some structures that are deemed worth simulating by the sampler. If there are resources available (GPUs an…
Figure 5
Figure 5. Figure 5: (A) A snapshot of a typical CG simulation showing a zoom in of the membrane associated RAS-RBDCRD protein complex. Water and ions are skipped for clarity, only the phosphate beads of the lipids are shown (cyan beads), and a backbone bead trace is shown for the protein …

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

68 extracted references · 67 canonical work pages

  1. [22]

    Stephany, Jeremy O

    Konstantia Georgouli, Robert R. Stephany, Jeremy O. B. Tempkin, Claudio Santiago, Fikret Aydin, Mark A. Heimann, Loïc Pottier, Xiaohua Zhang, Timothy S. Carpenter, Tim Hsu, Dwight V . Nissley, Frederick H. Streitz, Felice C. Lightstone, Helgi I. Ingolfsson, and Peer-Timo Bremer. Generating protein structures for pathway discovery using deep learning. Jour...

    work page doi:10.1021/acs.jctc.4c00816 2024

  2. [1]

    Top 500 supercomputers, 2024

    J Dongarra, H Meuer, and Erich Strohmaier. Top 500 supercomputers, 2024

    work page 2024

  3. [2]

    Exascale software study: Software challenges in extreme scale systems

    Saman Amarasinghe, Dan Campbell, William Carlson, Andrew Chien, William Dally, Elmootazbellah Elnohazy, Mary Hall, Robert Harrison, William Harrod, and Kerry Hill. Exascale software study: Software challenges in extreme scale systems. DARPA IPTO, Air Force Research Labs, Tech. Rep, pages 1–153, 2009

    work page 2009

  4. [3]

    Running many molecular dynamics simulations on many supercomputers

    Rajib Mukherjee, Abhinav Thota, Hideki Fujioka, Thomas C Bishop, and Shantenu Jha. Running many molecular dynamics simulations on many supercomputers. In Proceedings of the 1st Conference of the Extreme Science and Engineering Discovery Environment: Bridging from the eXtreme to the campus and beyond, pages 1–9, 2012

    work page 2012

  5. [4]

    Scaling of multimillion-atom biological molecular dynamics simulation on a petascale supercomputer

    Roland Schulz, Benjamin Lindner, Loukas Petridis, and Jeremy C Smith. Scaling of multimillion-atom biological molecular dynamics simulation on a petascale supercomputer. Journal of Chemical Theory and Computation, 5(10):2798–2808, 2009

    work page 2009

  6. [5]

    Hpc molecular simulation tries out a new gpu: Experiences on early amd test systems for the frontier supercomputer

    Ada Sedova, Russ Davidson, Mathieu Taillefumier, and Wael Elwasif. Hpc molecular simulation tries out a new gpu: Experiences on early amd test systems for the frontier supercomputer. Report, Oak Ridge National Lab.(ORNL), Oak Ridge, TN (United States), 2022

    work page 2022

  7. [6]

    Marrink, H

    Siewert J. Marrink, H. Jelger Risselada, Serge Yefimov, D. Peter Tieleman, and Alex H. de Vries. The martini force field: Coarse grained model for biomolecular simulations. The Journal of Physical Chemistry B, 111(27):7812– 7824, 2007. doi: 10.1021/jp071097f

    doi:10.1021/jp071097f 2007

  8. [7]

    Ingólfsson, Cesar A

    Helgi I. Ingólfsson, Cesar A. Lopez, Jaakko J. Uusitalo, Djurre H. de Jong, Srinivasa M. Gopal, Xavier Periole, and Siewert J. Marrink. The power of coarse graining in biomolecular simulations. WIREs Computational Molecular Science, 4(3):225–248, 2014

    work page 2014

Show all 68 references
  1. [8]

    100+ tflop solidification simulations on bluegene/l

    Frederick H Streitz, James N Glosli, Mehul V Patel, Bor Chan, Robert K Yates, Bronis R de Supinski, James Sexton, and John A Gunnels. 100+ tflop solidification simulations on bluegene/l. In Proceedings of IEEE/ACM Supercomputing, volume 5, 2005

    2005

  2. [9]

    J. N. Glosli, D. F. Richards, K. J. Caspersen, R. E. Rudd, J. A. Gunnels, and F. H. Streitz. Extending stability beyond cpu millennium: a micron-scale atomistic simulation of kelvin-helmholtz instability, 2007

    2007

  3. [10]

    Multiscale simulations of biological membranes: The challenge to understand biological phenomena in a living substance

    Giray Enkavi, Matti Javanainen, Waldemar Kulig, Tomasz Róg, and Ilpo Vattulainen. Multiscale simulations of biological membranes: The challenge to understand biological phenomena in a living substance. Chemical Reviews, 119(9):5607–5774, 2019. doi: 10.1021/acs.chemrev.8b00538

    2019 doi

  4. [11]

    Multiscale modelling and simulation: a position pa- per

    Alfons Hoekstra, Bastien Chopard, and Peter Coveney. Multiscale modelling and simulation: a position pa- per. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences , 372(2021):20130377, 2014. 10 Machine Learning-driven Multiscale MD...

    2021

  5. [12]

    Stansfeld and Mark S

    Phillip J. Stansfeld and Mark S. P. Sansom. From coarse grained to atomistic: A serial multiscale approach to membrane protein simulations. Journal of Chemical Theory and Computation, 7(4):1157–1166, 2011. doi: 10.1021/ct100569y

    2011 doi

  6. [13]

    Moon, Xiaohua Zhang, Joseph R

    Harsh Bhatia, Francesco Di Natale, Joseph Y . Moon, Xiaohua Zhang, Joseph R. Chavez, Fikret Aydin, Chris Stanley, Tomas Oppelstrup, Chris Neale, Sara Kokkila Schumacher, Dong H. Ahn, Stephen Herbein, Timothy S. Carpenter, Sandrasegaram Gnanakaran, Peer-Timo Bremer, James N. Gl...

    2021

  7. [14]

    Carpenter, Chris Neale, Sara Kokkila-Schumacher, Tomas Oppelstrup, Liam Stanton, Xiaohua Zhang, Shiv Sundram, Thomas R

    Francesco Di Natale, Harsh Bhatia, Timothy S. Carpenter, Chris Neale, Sara Kokkila-Schumacher, Tomas Oppelstrup, Liam Stanton, Xiaohua Zhang, Shiv Sundram, Thomas R. W. Scogland, Gautham Dharuman, Michael P. Surh, Yue Yang, Claudia Misale, Lars Schneidenbach, Carlos Costa, Cha...

    2019

  8. [15]

    Ingólfsson, Harsh Bhatia, Fikret Aydin, Tomas Oppelstrup, Cesar A

    Helgi I. Ingólfsson, Harsh Bhatia, Fikret Aydin, Tomas Oppelstrup, Cesar A. López, Liam G. Stanton, Timo- thy S. Carpenter, Sergio Wong, Francesco Di Natale, Xiaohua Zhang, Joseph Y . Moon, Christopher B. Stanley, Joseph R. Chavez, Kien Nguyen, Gautham Dharuman, Violetta Burns...

    2023

  9. [16]

    Multiscale scientific computation: Review 2001

    Achi Brandt. Multiscale scientific computation: Review 2001. Multiscale and Multiresolution Methods, pages 3–95. Springer Berlin Heidelberg, 2002

    2001

  10. [17]

    Hoekstra, and Joris Borgdorff

    Bastien Chopard, Jean-Luc Falcone, Alfons G. Hoekstra, and Joris Borgdorff. A framework for multiscale and multiscience modeling and numerical simulations. Unconventional Computation, pages 2–8. Springer Berlin Heidelberg, 2011

    2011

  11. [18]

    Roadmap on multiscale materials modeling

    Erik Van Der Giessen, Peter A Schultz, Nicolas Bertin, Vasily V Bulatov, Wei Cai, Gábor Csányi, Stephen M Foiles, Marc GD Geers, Carlos González, and Markus Hütter. Roadmap on multiscale materials modeling. Modelling and Simulation in Materials Science and Engineering, 28(4):0...

    2020

  12. [19]

    Ingólfsson, Chris Neale, Timothy S

    Helgi I. Ingólfsson, Chris Neale, Timothy S. Carpenter, Rebika Shrestha, Cesar A. López, Timothy H. Tran, Tomas Oppelstrup, Harsh Bhatia, Liam G. Stanton, Xiaohua Zhang, Shiv Sundram, Francesco Di Natale, Animesh Agarwal, Gautham Dharuman, Sara I. L. Kokkila Schumacher, Thomas...

    2022

  13. [20]

    Multiscale simulation of soft matter systems – from the atomistic to the coarse-grained level and back

    Christine Peter and Kurt Kremer. Multiscale simulation of soft matter systems – from the atomistic to the coarse-grained level and back. Soft Matter, 5(22):4357–4366, 2009

    2009

  14. [21]

    Carpenter, Felice C

    Harsh Bhatia, Fikret Aydin, Timothy S. Carpenter, Felice C. Lightstone, Peer-Timo Bremer, Helgi I. Ingólfsson, Dwight V . Nissley, and Frederick H. Streitz. The confluence of machine learning and multiscale simulations. Current Opinion in Structural Biology, 80:102569, 2023

    2023

  15. [23]

    F. Noé, S. Olsson, J. Köhler, and H. Wu. Boltzmann generators: Sampling equilibrium states of many-body systems with deep learning. Science, 365(6457), 2019. Export Date: 31 October 2024; Cited By: 405

    2019

  16. [24]

    Zuckerman and Lillian T

    Daniel M. Zuckerman and Lillian T. Chong. Weighted ensemble simulation: Review of methodology, applications, and software. Annual Review of Biophysics, 46(V olume 46, 2017):43–57, 2017

    2017

  17. [25]

    Carpenter, Helgi I

    Harsh Bhatia, Timothy S. Carpenter, Helgi I. Ingólfsson, Gautham Dharuman, Piyush Karande, Shusen Liu, Tomas Oppelstrup, Chris Neale, Felice C. Lightstone, Brian Van Essen, James N. Glosli, and Peer-Timo Bremer. 11 Machine Learning-driven Multiscale MD Workflows: The Mini-MuMM...

    2021

  18. [26]

    Exaworks: Workflows for exascale

    Aymen Al-Saadi, Dong H Ahn, Yadu Babuji, Kyle Chard, James Corbett, Mihael Hategan, Stephen Herbein, Shantenu Jha, Daniel Laney, and Andre Merzky. Exaworks: Workflows for exascale. In 2021 IEEE Workshop on Workflows in Support of Large-Scale Science (WORKS), pages 50–57. IEEE, 2021

    2021

  19. [27]

    Doe advanced scientific computing advisory subcommittee (ascac) report: top ten exascale research challenges

    Robert Lucas, James Ang, Keren Bergman, Shekhar Borkar, William Carlson, Laura Carrington, George Chiu, Robert Colwell, William Dally, and Jack Dongarra. Doe advanced scientific computing advisory subcommittee (ascac) report: top ten exascale research challenges. Report, USDOE...

    2014

  20. [28]

    Generalized-ensemble algorithms: enhanced sampling techniques for monte carlo and molecular dynamics simulations

    Yuko Okamoto. Generalized-ensemble algorithms: enhanced sampling techniques for monte carlo and molecular dynamics simulations. Journal of Molecular Graphics and Modelling, 22(5):425–439, 2004

    2004

  21. [29]

    Performance assessment of ensembles of in situ workflows under resource constraints

    Tu Mai Anh Do, Loïc Pottier, Rafael Ferreira da Silva, Silvina Caíno-Lores, Michela Taufer, and Ewa Deel- man. Performance assessment of ensembles of in situ workflows under resource constraints. Concurrency and Computation: Practice and Experience, 35(20):e7111, 2023

    2023

  22. [30]

    Elsa: An efficient, adaptive ensemble learning-based sampling approach

    Maria Böttcher, Alexander Fuchs, Ferenc Leichsenring, Wolfgang Graf, and Michael Kaliske. Elsa: An efficient, adaptive ensemble learning-based sampling approach. Advances in Engineering Software, 154:102974, 2021

    2021

  23. [31]

    Phillips, Klaus Schulten, and Christophe Chipot

    Jeffrey Comer, James C. Phillips, Klaus Schulten, and Christophe Chipot. Multiple-replica strategies for free- energy calculations in namd: Multiple-walker adaptive biasing force and walker selection rules. Journal of Chemical Theory and Computation, 10(12):5276–5285, 2014. do...

    2014 doi

  24. [32]

    Sumpter, Ondrej Dyck, Sergei V

    Ayana Ghosh, Bobby G. Sumpter, Ondrej Dyck, Sergei V . Kalinin, and Maxim Ziatdinov. Ensemble learning- iterative training machine learning for uncertainty quantification and automated experiment in atom-resolved microscopy. npj Computational Materials, 7(1):100, 2021

    2021

  25. [33]

    Ai-driven multiscale simulations illuminate mechanisms of sars-cov-2 spike dynamics

    Lorenzo Casalino, Abigail C Dommer, Zied Gaieb, Emilia P Barros, Terra Sztain, Surl-Hee Ahn, Anda Trifan, Alexander Brace, Anthony T Bogetti, Austin Clyde, Heng Ma, Hyungro Lee, Matteo Turilli, Syma Khalid, Lillian T Chong, Carlos Simmerling, David J Hardy, Julio DC Maia, Jame...

    2021

  26. [34]

    Transparent in situ data transformations in adios

    David A Boyuka, Sriram Lakshminarasimham, Xiaocheng Zou, Zhenhuan Gong, John Jenkins, Eric R Schendel, Norbert Podhorszki, Qing Liu, Scott Klasky, and Nagiza F Samatova. Transparent in situ data transformations in adios. In 2014 14th IEEE/ACM International Symposium on Cluster...

    2014

  27. [35]

    In-situ processing and visualization for ultrascale simulations

    Ma Kwan-Liu, Wang Chaoli, Yu Hongfeng, and Tikhonova Anna. In-situ processing and visualization for ultrascale simulations. Journal of Physics: Conference Series, 78(1):012043, 2007

    2007

  28. [36]

    In situ methods, infrastructures, and applications on high performance computing platforms

    Andrew C Bauer, Hasan Abbasi, James Ahrens, Hank Childs, Berk Geveci, Scott Klasky, Kenneth Moreland, Patrick O’Leary, Venkatram Vishwanath, and Brad Whitlock. In situ methods, infrastructures, and applications on high performance computing platforms. In Computer Graphics Foru...

    2016

  29. [37]

    Accelerating scientific workflows on hpc platforms with in situ processing

    Tu Mai Anh Do, Loïc Pottier, Orcun Yildiz, Karan Vahi, Patrycja Krawczuk, Tom Peterka, and Ewa Deelman. Accelerating scientific workflows on hpc platforms with in situ processing. In 2022 22nd IEEE International Symposium on Cluster, Cloud and Internet Computing (CCGrid), page...

    2022

  30. [38]

    Combining in-situ and in-transit processing to enable extreme-scale scientific analysis

    Janine C Bennett, Hasan Abbasi, Peer-Timo Bremer, Ray Grout, Attila Gyulassy, Tong Jin, Scott Klasky, Hemanth Kolla, Manish Parashar, and Valerio Pascucci. Combining in-situ and in-transit processing to enable extreme-scale scientific analysis. In SC’12: Proceedings of the Int...

    2012

  31. [39]

    Ensemble toolkit: Scalable and flexible execution of ensembles of tasks

    Vivekanandan Balasubramanian, Antons Treikalis, Ole Weidner, and Shantenu Jha. Ensemble toolkit: Scalable and flexible execution of ensembles of tasks. In 2016 45th International Conference on Parallel Processing (ICPP), pages 458–463. IEEE, 2016

    2016

  32. [40]

    Enabling machine learning-ready hpc ensembles with merlin

    J Luc Peterson, Ben Bay, Joe Koning, Peter Robinson, Jessica Semler, Jeremy White, Rushil Anirudh, Kevin Athey, Peer-Timo Bremer, and Francesco Di Natale. Enabling machine learning-ready hpc ensembles with merlin. Future Generation Computer Systems, 131:255–268, 2022

    2022

  33. [41]

    Tran, Albert H

    Timothy H. Tran, Albert H. Chan, Lucy C. Young, Lakshman Bindu, Chris Neale, Simon Messing, Srisathiya- narayanan Dharmaiah, Troy Taylor, John-Paul Denson, Dominic Esposito, Dwight V . Nissley, Andrew G. Stephen, Frank McCormick, and Dhirendra K. Simanshu. Kras interaction wit...

    2021

  34. [42]

    D. K. Simanshu, D. V . Nissley, and F. McCormick. Ras proteins and their regulators in human disease. Cell, 170(1):17–33, 2017. 1097-4172 Simanshu, Dhirendra K Nissley, Dwight V Mc- Cormick, Frank HHSN261200800001C/CA/NCI NIH HHS/United States HHSN261200800001E/CA/NCI NIH HHS/...

    2017 doi

  35. [43]

    A. M. Waters and C. J. Der. Kras: The critical driver and therapeutic target for pancreatic cancer. Cold Spring Harb Perspect Med, 8(9), 2018. 2157-1422 Waters, Andrew M Der, Channing J R01 CA042978/CA/NCI NIH HHS/United States T32 CA009156/CA/NCI NIH HHS/United States P01 CA2...

    2018

  36. [44]

    Tackling exascale software challenges in molecular dynamics simulations with gromacs

    Szilárd Páll, Mark James Abraham, Carsten Kutzner, Berk Hess, and Erik Lindahl. Tackling exascale software challenges in molecular dynamics simulations with gromacs. In Solving Software Challenges for Exascale: International Conference on Exascale Applications and Software, EA...

    2014

  37. [45]

    Carpenter, Tomas Oppelstrup, Xiaohua Zhang, Peer-Timo Bremer, Fred Streitz, Felice C

    Fikret Aydin, Konstantia Georgouli, Loïc Pottier, Timothy S. Carpenter, Tomas Oppelstrup, Xiaohua Zhang, Peer-Timo Bremer, Fred Streitz, Felice C. Lightstone, and Helgi I. Ingólfsson. Enhancing exploration of protein orientational space through integration of ultra-coarse-grai...

    2024 doi

  38. [46]

    Klepeis, Kresten Lindorff-Larsen, Ron O

    John L. Klepeis, Kresten Lindorff-Larsen, Ron O. Dror, and David E. Shaw. Long-timescale molecular dynamics simulations of protein structure and function. Current Opinion in Structural Biology, 19(2):120–127, 2009

    2009

  39. [47]

    Lammps-a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales

    Aidan P Thompson, H Metin Aktulga, Richard Berger, Dan S Bolintineanu, W Michael Brown, Paul S Crozier, Pieter J In’t Veld, Axel Kohlmeyer, Stan G Moore, and Trung Dac Nguyen. Lammps-a flexible simulation tool for particle-based materials modeling at the atomic, meso, and cont...

    2022

  40. [48]

    Lyman, J

    E. Lyman, J. Pfaendtner, and G. A. V oth. Systematic multiscale parameterization of heterogeneous elastic network models of proteins. Biophys J, 95(9):4183–92, 2008. 1542-0086 Lyman, Edward Pfaendtner, Jim V oth, Gregory A R01 GM063796/GM/NIGMS NIH HHS/United States Journal Ar...

    2008 doi

  41. [49]

    Zhang, L

    Z. Zhang, L. Lu, W. G. Noid, V . Krishna, J. Pfaendtner, and G. A. V oth. A systematic methodology for defining coarse-grained sites in large biomolecules. Biophys J, 95(11):5073–83, 2008. 1542-0086 Zhang, Zhiyong Lu, Lanyuan Noid, Will G Krishna, Vinod Pfaendtner, Jim V oth, ...

    2008

  42. [50]

    Oppelstrup, L

    T. Oppelstrup, L. G. Stanton, J. O. B. Tempkin, T. Ozturk, H. I. Ingólfsson, and T. S. Carpenter. Anisotropic interactions for continuum modeling of protein-membrane system. The Journal of Chemical Physics, 2024

    2024

  43. [51]

    L. G. Stanton, T. Oppelstrup, T. S. Carpenter, H. I. Ingólfsson, M. P. Surh, F. C. Lightstone, and J. N. Glosli. Dy- namic density functional theory of multicomponent cellular membranes. Physical Review Research, 5(1):013080,

  44. [52]

    Paulo C. T. Souza, Riccardo Alessandri, Jonathan Barnoud, Sebastian Thallmair, Ignacio Faustino, Fabian Grünewald, Ilias Patmanidis, Haleh Abdizadeh, Bart M. H. Bruininks, Tsjerk A. Wassenaar, Peter C. Kroon, Josef Melcr, Vincent Nieto, Valentina Corradi, Hanif M. Khan, Jan Do...

    2021

  45. [53]

    Ingólfsson, Harsh Bhatia, Talia Zeppelin, W

    Helgi I. Ingólfsson, Harsh Bhatia, Talia Zeppelin, W. F. Drew Bennett, Kristy A. Carpenter, Pin-Chia Hsu, Gautham Dharuman, Peer-Timo Bremer, Birgit Schiøtt, Felice C. Lightstone, and Timothy S. Carpenter. Capturing biologically complex tissue-specific membranes at different l...

    2020 doi

  46. [54]

    T. N. Ozturk, M. König, T. S. Carpenter, K. B. Pedersen, T. A. Wassenaar, H. I. Ingólfsson, and S. J. Mar- rink. Building complex membranes with martini 3. Methods Enzymol, 701:237–285, 2024. 1557-7988 Oz- turk, Tugba Nur König, Melanie Carpenter, Timothy S Pedersen, Kasper B ...

    2024 doi

  47. [55]

    Wassenaar, Helgi I

    Tsjerk A. Wassenaar, Helgi I. Ingólfsson, Rainer A. Böckmann, D. Peter Tieleman, and Siewert J. Marrink. Com- putational lipidomics with insane: A versatile tool for generating custom membranes for molecular simulations. Journal of Chemical Theory and Computation, 11(5):2144–2...

    2015 doi

  48. [56]

    RabbitMQ in action: distributed messaging for everyone

    Jason Williams. RabbitMQ in action: distributed messaging for everyone. Simon and Schuster, 2012

    2012

  49. [57]

    Maestro workflow conductor

    Francesco Di Natale. Maestro workflow conductor. Report, Lawrence Livermore National Laboratory (LLNL), Livermore, CA (United States), 2017

    2017

  50. [58]

    Slurm: Simple linux utility for resource management

    Andy B Yoo, Morris A Jette, and Mark Grondona. Slurm: Simple linux utility for resource management. In Workshop on job scheduling strategies for parallel processing, pages 44–60. Springer, 2003

    2003

  51. [59]

    Scalable, fault-tolerant job step management for high performance systems

    D Solt, J Hursey, A Lauria, D Guo, and X Guo. Scalable, fault-tolerant job step management for high performance systems. 2019

    2019

  52. [60]

    Flux: Overcoming scheduling challenges for exascale workflows

    Dong H Ahn, Ned Bass, Albert Chu, Jim Garlick, Mark Grondona, Stephen Herbein, Helgi I Ingólfsson, Joseph Koning, Tapasya Patki, and Thomas RW Scogland. Flux: Overcoming scheduling challenges for exascale workflows. Future Generation Computer Systems, 110:202–213, 2020

    2020

  53. [61]

    Resilience-aware resource management for exascale computing systems

    Daniel Dauwe, Sudeep Pasricha, Anthony A Maciejewski, and Howard Jay Siegel. Resilience-aware resource management for exascale computing systems. IEEE Transactions on Sustainable Computing, 3(4):332–345, 2018

    2018

  54. [62]

    The spack package manager: bringing order to hpc software chaos

    Todd Gamblin, Matthew LeGendre, Michael R Collette, Gregory L Lee, Adam Moody, Bronis R De Supinski, and Scott Futral. The spack package manager: bringing order to hpc software chaos. In Proceedings of the International Conference for High Performance Computing, Networking, St...

    2015

  55. [63]

    Streitz, James N

    Frederick H. Streitz, James N. Glosli, and Mehul V . Patel. Beyond finite-size scaling in solidification simulations. Physical Review Letters, 96(22):225701, 2006. PRL

    2006

  56. [64]

    Xiaohua Zhang, Shiv Sundram, Tomas Oppelstrup, Sara I. L. Kokkila-Schumacher, Timothy S. Carpenter, Helgi I. Ingólfsson, Frederick H. Streitz, Felice C. Lightstone, and James N. Glosli. ddcmd: A fully gpu-accelerated molecular dynamics program for the martini force field. The ...

    2020

  57. [65]

    Redis in Action

    J Carlson. Redis in Action. Manning, 2013

    2013

  58. [66]

    Container utilization at doe compute facilities

    Andrew J Younge, Anthony Michael Agelastos, and Kevin Pedretti. Container utilization at doe compute facilities. Report, Sandia National Lab.(SNL-NM), Albuquerque, NM (United States), 2020

    2020

  59. [67]

    O’Reilly Media, Inc

    Brendan Burns, Joe Beda, Kelsey Hightower, and Lachlan Evenson. Kubernetes: up and running. " O’Reilly Media, Inc.", 2022

    2022

  60. [68]

    One step closer to converged computing: Achieving scalability with cloud-native hpc

    Daniel J Milroy, Claudia Misale, Giorgis Georgakoudis, Tonia Elengikal, Abhik Sarkar, Maurizio Drocco, Tapasya Patki, Jae-Seung Yeom, Carlos Eduardo Arango Gutierrez, and Dong H Ahn. One step closer to converged computing: Achieving scalability with cloud-native hpc. In 2022 I...

    2022

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.