DireX - Tutorial
Kinked Helix at High Resolution
In this tutorial the focus is on sidechain placement.
The starting structure is the same extended helix as in the previous low-resolution example.
This test case is constructed such that the sidechains can be in significantly different
orientations than in the starting structure.
Initial structure (orange) and target density (green).
Clearly a 3 Å density map (and this ideal map in particular) defines the structure
and also most sidechains quite well.
We therefore switch off the DEN restraints
on the sidechains which is achieved by setting den_sidechain
to zero. This parameter defines a factor
that is multiplied to the strength of those DEN restraints that involve sidechain
You will immediately realize that running this case takes longer than the low-resolution
case. There are two reasons:
(1) density map has more grid points which makes the computation slightly more costly
and (2) the sampling takes longer since the 'energy landscape' (the fit function)
is more rugged and has many more local minima that need to be crossed. Finding the correct
optimum becomes more difficult. In some cases it might even be advisable to filter the
density map to lower a resolution in a first round of refinement, and then increase the resolution
in subsequent rounds until the highest resolution is reached.
The refinement is performed here in two steps:
The first DireX run samples more aggressively to cross 'energy' barriers and the second run
is more like
a minimization (although not in a strict sense, remember that DireX does not use
gradients to move atoms but instead uses a geometry-based sampling algorithm).
The first DireX run:
# Run refinement
direx -f refine.par -pdb extended-helix.pdb -cur current-refine.pdb -curmap curmap.mrc -map kinked-density.mrc -ox traj.xtc
and the second (subsequent) run uses the output structure ('current-refine.pdb
') from the
# Run minimization
direx -f min.par -pdb extended-helix.pdb -p current-refine.pdb -cur current-min.pdb -curmap curmap-min.mrc -map kinked-density.mrc -ox traj-min.xtc
Fitted structure (blue) and target density (green grid).
Close-up view of fitted structure (blue), target (correct) structure (green) and target density (green grid).
The topology/geometry (bonds and angles, etc) should always be read (with -pdb
) from the
initial structure ('extended-helix
Since DireX allows for (very small) deviations of bond lengths and angles, such small errors could
accumulate if the output structure is used as input several times. The actual coordinates are then
read with the -p
Look at the resulting trajectories with VMD:
vmd -f extended-helix.pdb traj.xtc traj-min.xtc -m kinked-density.mrc target.pdb
Reading the target structure as well helps to evaluate the final fitted structure. You might
also want to read in curmap.mrc
to compare it to the target map kinked-density.mrc
As an exercise, you can switch off the DEN restraints (set use_den = no
You will find that the DEN restraints will help to guide the refinement towards
the optimal structure even at this (relatively) high resolution.