Describe it
Say what you want in a sentence. The Agent writes the study, and the same validator that checks every study checks it before anything runs.
A two-dimensional Lennard-Jones fluid, simulated in your browser just below its critical point. Droplets of 10 or more atoms are drawn in teal.
Early access is open
From question to publishable molecular dynamics
Every study is reproducible, traceable and checked before it runs.
MDXplora prepares the system, runs the dynamics, analyses the trajectory and writes up the study, in your browser.
Simulate trypsin with benzamidine bound at pH 6.5 for 100 ns
ConfigValidated
systems:
- system: 3PTB
setup:
ph: 6.5
forcefield: amber-openff
ligand_name: BEN
simulation:
duration_ns: 100Run onMDXplora GPUYour workstation
How a study runs
A Config goes in. A Manifest comes out.
One file describes the whole study. Four phases run from it, each in its own directory, and one record says what actually happened.
- 00
the studyConfigOne file: the system, how it is prepared, how it is simulated, what is measured and how it is reported.
- 01
phasesetupFetches or reads the structure, repairs what is missing, decides what each non-standard residue is for, then solvates and adds ions.
- 02
phasesimulationMinimisation, NVT and NPT equilibration, then production. Restraints, membranes and biased sampling when the study asks for them.
- 03
phaseanalysisEach analysis measures its own copy of the trajectory and records the settings it actually used.
- 04
phasereportA methods paragraph assembled from recorded values, and a convergence section that says what the trajectory can support.
- 05
the recordManifestEvery phase, artifact and setting, and the software stack it ran on. Readable by someone who was not there.
Three ways to start
Describe it, build it, or bring it.
Build it
A form with every setting visible and explained, checked as you go, for when you know exactly what you want.
Bring it
Upload a Config you already have. It runs as written, and every study can be downloaded as a Config again.
What you can study
From one protein to a campaign.
A protein on its own
Fold, flexibility, secondary structure, native contacts and conformational clustering, from a PDB code. Fluctuations can be set against the crystal’s own B-factors.
A protein with a ligand
The ligand is found, its chemistry resolved and its protonation settled in the binding site. Interactions against published criteria show what holds it, not only what it touches.
A membrane protein
Embedded in a lipid bilayer, with its orientation checked rather than assumed and pressure coupling suited to a lipid system.
Free energy along a coordinate
Umbrella sampling, metadynamics and steered MD from a named collective variable. Each result says what it is and what it is not.
A trajectory you already have
Skip the simulation and analyse trajectories written by other engines, in their own formats.
Many systems at once
Mutants against wild type, a sweep across a setting, one run per GPU, and a comparison report across all of them.
Scientific integrity
It refuses rather than guesses.
An ambiguous ligand charge, a protein backwards in its membrane, a free-energy surface that never converged: each stops the run and is named, not papered over.
- Every refusal carries a sentence for a person and a code for a program.
- Biased averages are corrected to equilibrium where the bias allows, and labelled where it does not.
- A standard error the trajectory cannot support is withheld, not printed beside a caveat.
!Setup stopped
Ligand BEN: the study states a net charge of +1, and the chemistry file carries formal charges summing to 0. The file is the chemistry, so the two cannot both stand.
Supply the ligand in the protonation state the study means, then run again.
Compute
Our GPUs, or yours.
Run on GPUs we provide, or connect your own workstation or cluster in one command. Your machine asks MDXplora for work, so nothing reaches into your network.
How your own compute connects →Your data stays yours
Every study can be downloaded whole: the Config, the data, the report and the Manifest.
Services
When you would rather we ran it.
For groups without someone to run molecular dynamics, and for labs that depend on it.
Services →Studies run for you
You get the report, the data and the full record behind them.
Training for your group
From a first study to your own campaigns, on your own systems.
Support
For your machines, for studies that refuse, and for results that need a second pair of eyes.
Questions
Before you ask.
Can I use MDXplora today?
MDXplora is in early access. Join the list and we will be in touch as places open, starting with a small number of groups.
Do I need a GPU?
No. Run on GPUs we provide, or connect your own workstation or cluster and choose for each study where it runs.
Where does my data go?
Studies run on our GPUs are stored for your account. Studies run on your own machines can keep their trajectories on your own storage, with only the records and reports shown in MDXplora.
How do I cite results?
Results produced on MDXplora are computed with FastMDXplora. Cite its paper (10.1002/jcc.70350) and the release named in your study’s Manifest (10.5281/zenodo.17510591).