DiPhyxDiPhyx
Workspace
The name

It is calledenergy minimisation

People say relaxation, and everyone knows what they mean. The manual says minimisation, and so do the input files.

A freshly built system is wrong in small, violent ways. Two atoms sit slightly too close, a side chain overlaps a water molecule, a hydrogen was placed by a script that did not know better. Left alone, the first integration step turns each of those into an enormous force, and the run ends in the log file rather than in your results.

Steric clashes

Atoms placed closer than their radii allow, which the Lennard-Jones term reads as a near-infinite force.

Added hydrogens

Crystal structures rarely include them, so they are placed geometrically and are approximately wrong.

Fresh solvent

A water box dropped around a solute leaves molecules in positions no equilibrated liquid would have.

The sequence

Four short runsbefore the long one

Each stage removes one kind of wrongness, and each is cheap compared with the production run it protects.

MinimiseDownhill in potential energy at 0 K, usually twice: solvent first with the solute held, then everything.
HeatNVT, from near zero to your target temperature over tens of picoseconds, with the solute still restrained.
Equilibrate densityNPT, so the box settles to the right volume. Watch density approach roughly 1 g/cm³ for a water system.
Release, then produceEase the restraints off over a few short runs, then start the production trajectory you will analyse.

Minimisation happens effectively at zero kelvin — there is no kinetic energy, only a walk downhill in potential energy using steepest descent and then conjugate gradient. Temperature only enters once heating begins, which is why minimisation cannot fix a structure that is wrong in a way that needs motion to escape.

Restraints

Hold the partsyou already trust

Position restraints are the difference between relaxing the water around your protein and quietly relaxing your protein into something else.

Hold the solute

Restrain heavy atoms while solvent and hydrogens relax around a structure you already believe in.

Let go gradually

Drop the force constant in stages rather than all at once, so nothing is released into a shock.

Keep what is fixed, fixed

Crystallographic waters, a metal centre, a covalent modification — restrain them or lose them.

Knowing it worked

Two numbersand one picture

Minimisation converging is not the same as the system being ready. Check the obvious things before committing machine hours.

Falls flatEnergysteep, then a plateau
SmallMax forceno single atom left straining
Look at itThena converged failure still converges

Watch the energy fall and flatten, watch the maximum force per atom drop to something small, and then look at the structure. A minimisation that converged beautifully on a system where the ligand drifted out of the pocket is a converged failure, and no number in the output will tell you that.

The status

AmberMD isbeing built

The image and its definition are being verified. When they are done, it deploys by name like the rest.

Ask for itSay so on the community tracker, and it moves up the list.Or build itOpen a pull request on the hub, following a published one.
Read it, then run itPrepaid and by the hour, on a machine that is yours about a minute after you ask.
Start a machineAll notes