Alanine dipeptide¶

Tested against md-tools 0.6.1. The smallest system in this set: two peptide bonds capped at both ends, 22 atoms, and the standard toy for everything that depends on backbone torsions.

The alanine dipeptide

ACE–ALA–NME. Its two backbone torsions, φ and ψ, are the coordinates every method on this page is measured against.

Why this system¶

φ and ψ have barriers of a few kT, so a plain simulation crosses them often enough to give a reference distribution in minutes rather than days. That makes it the one system where an enhanced method can be checked against the truth instead of against another enhanced method.

Build it¶

The structure ships with the package as tests/data/ALA.pdb, and the build is one command. For an explicit box:

# build.config -- explicit solvent
solvent: {model: TIP3P, padding_nm: 1.0}

For implicit solvent, which for a system this small is often enough and has no box at all:

# build.config -- implicit solvent
solvent: {model: GBn2}

md-openmm build-top -i ALA.pdb --config build.config \
    -os build/built.xml -op build/built.pdb -log build/built.log

An implicit build has no box, no barostat, no counter-ions and no NPT stage; the stages are renamed accordingly rather than run at a pressure that means nothing.

Where to go next¶

method	what it is for
cMD	the unbiased reference every method here is checked against, 10 ns	published
REST2	solute tempering, four states over τ 0 → 0.5	published
AIS	annealed importance sampling between two Hamiltonians, with the torsion reweighting checked against the plain simulation	published
Umbrella sampling	a potential of mean force along φ	page planned, 0.6.5
Alchemical — ligand to ligand	free energies by transformation	planned, 0.6.5

Umbrella sampling arrives in 0.6.3 and the alchemical pages in 0.7.0. They are listed here so the shape of the set is visible; neither is written yet, and neither is linked to a page that does not exist.