PaperlayerView exact revision on arXiv

Source-linked AI summary

Sampling molecular conformations and dynamics in a multi-user virtual reality framework

Michael O Connor, Helen M. Deeks, Edward Dawn, Oussama Metatla, Anne Roudaut, Matthew Sutton, Becca Rose Glowacki, Rebecca Sage, Philip Tew, Mark Wonnacott, Phil Bates, Adrian J. Mulholland, David R. Glowacki

arXiv:1801.02884v1physics.chem-phcs.HCphysics.bio-phphysics.ed-ph

TL;DR

Three-dimensional co-location is a key challenge for interactive molecular simulation, where molecules move in three dimensions. The paper combines commodity VR hardware with cloud-based molecular dynamics and reports faster task completion than conventional interfaces, with the advantage appearing to increase as task complexity rises.

  • Problem

    Interactive molecular simulations face a 3D co-location challenge because molecules and their motion are three-dimensional, while conventional interfaces have limitations in aligning physical and virtual interaction sites.

    Source: p. 2

  • Method

    The framework combines tracked commodity VR hardware with real-time molecular dynamics, using cloud computing to access more powerful simulation servers as needed.

    Source: p. 2

  • Results

    The VR platform showed a clear advantage over conventional visualization and interaction platforms, and that advantage appeared to increase with problem complexity.

    Source: p. 4

  • Takeaways & Limitations

    For biomolecular conformational problems such as dynamical path sampling or drug docking, VR frameworks may offer a powerful tool for research experts.

    Source: p. 4

Abstract

from arXiv · show

We describe a framework for interactive molecular dynamics in a multiuser virtual reality environment, combining rigorous cloud-mounted physical atomistic simulation with commodity virtual reality hardware, which we have made accessible to readers (see isci.itch.io/nsb-imd). It allows users to visualize and sample, with atomic-level precision, the structures and dynamics of complex molecular structures 'on the fly', and to interact with other users in the same virtual environment. A series of controlled studies, wherein participants were tasked with a range of molecular manipulation goals (threading methane through a nanotube, changing helical screw-sense, and tying a protein knot), quantitatively demonstrate that users within the interactive VR environment can complete sophisticated molecular modelling tasks more quickly than they can using conventional interfaces, especially for molecular pathways and structural transitions whose conformational choreographies are intrinsically 3d. This framework should accelerate progress in nanoscale molecular engineering areas such as drug development, synthetic biology, and catalyst design. More broadly, our findings highlight VR's potential in scientific domains where 3d dynamics matter, spanning research and education.

Supplementary Information

The Supplementary Information lists study and simulation resources and includes the paper’s author and affiliation details.

1. Molecular Dynamics Simulations

The supplement outlines interactive molecular dynamics as a user-guided modification of classical atomistic simulation. It separates internal molecular energy from the energy of the user’s interaction.

  • The supplement briefly summarizes interactive-MD strategies previously outlined in reference (1).Source: §1 · p. 7
  • Classical molecular dynamics numerically integrates Newton’s equations using forces derived from the system’s potential energy.Source: §1 · p. 7
  • The system lets users chaperone real-time molecular dynamics by splitting potential energy into internal and user-exerted components.Source: §1 · p. 7

Eq (S3) then gives

The supplement describes how user interaction is applied as an external force and details the simulation settings and study hardware used for the tasks.

  • For all tasks on all platforms, external forces were implemented by projecting a spherical Gaussian field into the system, with its strength set to 2000 (kJ*mol-1)/(a.m.u.).This value was chosen to achieve responsive interaction while preserving dynamical stability.Source: Eq (S3) then gives · pp. 7–8
  • While an interaction is active, it stays on the same atom, and users can adjust its course and strength by repositioning the field.Source: Eq (S3) then gives · p. 8
  • The mouse and touchscreen interaction field attaches to the nearest atom in the screen plane, so 3D motion requires successive 2D movements and camera repositioning.In VR, the field is co-located with the controller.Source: Eq (S3) then gives · p. 8
  • The nanotube and helicene tasks used an in-house MM3 forcefield, while the 17-Alanine protein-knot task used OpenMM’s GPU-accelerated Amber99SB-ILDN forcefield; all tasks used a Velocity Verlet integrator.All tasks used a Berendsen thermostat set to 200K. The nanotube and helicene tasks used a 1fs time step, versus 2fs for the protein-knot task.Source: Eq (S3) then gives · p. 8
  • The study used separate local-area-network machines for each task to avoid latency from excessive computational load on one machine.Source: Eq (S3) then gives · p. 8
  • All platforms were throttled to 30 FPS to provide an equally fluid user experience despite differences in machine specifications.Source: Eq (S3) then gives · p. 9

2. Launching a cloud hosted interactive MD session

The supplement notes that executables were provided for multiple platforms.

  • Binaries were provided for Windows desktop, Windows with SteamVR, and Mac OS X.Source: §2 · p. 10

Android) are available at https://isci.itch.io/nsb-imd

This implementation section explains how to launch the cloud-hosted framework on supported devices and how server distance and network conditions affect latency. It reports measurements across three cloud centers and practical access constraints.

Android) are available at https://isci.itch.io/nsb-imd

  • The application supports Windows and Mac desktops or laptops, Android touchscreen devices, and Windows 10 VR desktops; iOS is not supported.The VR version requires a VR-capable GPU, and the listed Android devices have additional processor and graphics requirements.Source: Android) are available at… · p. 10
  • Latency depends on network speed, wifi performance, and distance from the server, so the instructions recommend selecting the nearest available Oracle data center.The listed centers are Frankfurt, Phoenix, and Ashburn.Source: Android) are available at… · p. 10
  • 48.7 ± 9.5 ms to Frankfurt, 118.7 ± 11.5 ms to Ashburn, and 182.1 ± 18.3 ms to Phoenix were the average Bristol round-trip latencies.The measurements report average latency ± standard deviation.Source: Android) are available at… · p. 10
  • Wifi and ethernet latencies were statistically indistinguishable, and acceptable latencies were also observed on 4G mobile networks.Figure S1 compares wifi in its top panel with ethernet in its bottom panel.Source: Android) are available at… · p. 10, Figure S1 · p. 11
  • User-study simulation servers ran on a local network because Oracle's Frankfurt cloud resources, which had the best Bristol latencies, were not yet available.Source: Android) are available at… · p. 11
  • The cloud-mounted engine offers readers a straightforward way to test the framework and avoids potential hardware and operating-system compatibility issues.Readers unable to obtain reasonable latency on available cloud servers may contact the corresponding author for a testing license.Source: Android) are available at… · p. 11

3. User Study Design

The section describes three user studies evaluating molecular-manipulation task designs across VR, screen-and-mouse, and touchscreen interfaces, including knot-task evaluation. It covers participants’ preferences and reported advantages and difficulties with the different controls for perceiving and positioning molecules in 3D.

3.1 First User Study

  • The first study enrolled 32 participants who tried mouse, touchscreen, and VR in randomized Latin-square order, practicing on buckminsterfullerene before nanotube/methane and helicene tasks.Participants completed a questionnaire after attempting both tasks on all three platforms.Source: §3.1 · p. 12

3.2 Second User Study

  • The second study used 12 participants to gather qualitative feedback on three platforms through interviews and thematic analysis after a trefoil-knot task.Participants also completed the first study’s questionnaires, and watched an instructional video before attempting the knot.Source: §3.2 · pp. 13–14

3.3 Third User Study

  • The third study repeated the knot-task methodology with 32 participants to improve completion statistics, with no questionnaire or interview afterward.Platform starting order was randomized and participants rotated through platforms using a Latin square.Source: §3.3 · p. 15

3.4 Outline of User Tasks

  • The introductory buckminsterfullerene task let users practice grabbing, moving, resizing, and rotating molecules and the virtual space, with no time limit or specific endpoint.Source: §3.4 · p. 16
  • Participants manipulated methane through a nanotube, reversed a helicene’s screw-sense, and tied a trefoil knot in 17-alanine.The tasks had time limits of 180 seconds for nanotube threading and knot tying, and 210 seconds for helicene reversal.Source: §3.4 · pp. 16–17

3.5 Platform controls

  • VR users could walk around a molecule or rotate the 3D camera; two grip buttons scaled the render, controller triggers grabbed atoms, and 3D motions required successive 2D motions with camera repositioning.Source: §3.5 · pp. 17, 19

3.5 Thematic Analysis

  • 41 of 44 participants preferred VR, saying its depth perception made molecular 3D shapes easier to understand than on flat-screen platforms.Participants also reported that it was harder to see a molecule’s overall shape on the two flat-screen platforms.Source: §3.5 · p. 21
  • Every questionnaire response rated depth perception important to some degree, the most positive response; 32 task participants completed questionnaires and 12 knot-task participants were interviewed in groups.The questionnaires followed the nanotube/methane and helicene tasks; the interviews followed the knot task.Source: §3.5 · p. 21
  • VR’s immersive fully-3D space let users easily move into and out of the screen plane, whereas participants using flat-screen mouse and touchscreen platforms found pinpointing 3D space more difficult.Source: §3.5 · p. 22
  • During knot tying on flat screens, participants changed camera orientation to inspect the loop, temporarily surrendering molecular control and sometimes losing the molecule’s shape.VR allowed users to control the camera and molecule simultaneously and manipulate molecules in any dimension.Source: §3.5 · p. 22
  • Touchscreen users sometimes rotated the camera when intending to grab a molecule, whereas VR separated head-controlled viewing from hand-controlled molecular manipulation.The shared touchscreen drag action created input ambiguity; separate VR controls removed it.Source: §3.5 · p. 22
  • Participants found VR interactions more intuitive because familiar physical gestures mapped directly to molecular actions, including tying a virtual knot like a physical one.On the other platforms, participants translated familiar gestures into less intuitive secondary gestures, particularly for complicated 3D tasks.Source: §3.5 · p. 23

4. Web URLs for Supplementary Videos

The section provides links to two supplementary videos.

4. Web URLs for Supplementary Videos

  • Supplementary Video 1 is available at https://vimeo.com/244670465.Source: §4 · p. 23
  • Supplementary Video 2 is available at https://vimeo.com/235894288.Source: §4 · p. 23
  • Both supplementary video links point to Vimeo.Source: §4 · p. 23
Loading 1801.02884v1…