Apart from the standard procedures for detecting ligands (difference Fourier maps), BUSTER has one particular feature that needs a bit further explanation - to explain what it can do, what it can't do and what potential bias it might introduce.

This feature is triggered by the –L and –Lpdb flags to the refine command:

% refine -p some.pdb -m some.mtz -L ...
  - or -
% refine -p some.pdb -m some.mtz -Lpdb bindingsite.pdb ...

It will treat a certain region of the model differently during the last big cycle of refinement: that region will neither contain an atomic model nor a contribution from the bulk solvent. However, if there is some electron density present in this region, it should show up in difference (Fo-Fc) maps as strong positive density. The interpretation, what this electron density might represent (atomic model, bulk solvent or a mixture) is up to the user.

Here we're going to explain the typical usage of this BUSTER feature, their assumptions and caveats.

Unknown location

The least biased assumption is that there might be some ligand bound, but its location is unknown. In that case one would use the

% refine -p some.pdb -m some.mtz -L ...

option. It will also trigger the automatic addition of 'dummy' atoms (currently described in terms of 'water' residues). Just before the last big cycle, an analysis of regions badly explained by those dummy atoms is done. This could result in a report (on standard output) about a certain number of waters being removed in a so-called 'interesting' region. This region is then treated as described above.

Known location

If the location of ligand binding is known, a PDB file describing this region can be used explicitly with

% refine -p some.pdb -m some.mtz -Lpdb bindingsite.pdb ...

The same logic as before will be applied just before the last big cycle - with one difference: instead of searching for regions with unexplained difference density, the region described by the user-given PDB file is being used.


Some points need to be considered here:

1. the placed 'waters'

Since the goal of this procedure is to find, highlight and enhance one or more (potential) ligand-binding sites, the placed 'waters' need to be taken not too serious. They are placed to explain density, which could be non-water density (missing loops, disordered side-chains, unmodelled compounds) but also just noise in the data. It is not recommended to use the resulting solvent structure as such.

2. false negatives

It is possible to miss genuine bound compounds - especially if they are weakly bound (partial occupancy and/or partially disordered) or rather small (fragment searches). There are options available for changing the minimum required volume.

3. false positives

It is possible to get false positives with that procedure - especially the –Lpdb flag. Some typical examples are:

  • good high resolution data will start showing more complicated water clusters with partially occupied, alternate conformations for ordered waters. These can easily be mistaken as a connected compound by the current implementation (but they should also be easily recognisable when checking the difference maps).
  • unmodelled macromolecule (loops, N- or C-terminii etc) could show up as larger, connected regions. It is recommended to repeat the procedure after correcting those issues within the atomic model of the macromolecule (which will also improve the phases and therefore the difference density)
  • using a description of the binding site (with the –Lpdb option) that corresponds very closely to the shape of the assumed compound bound in that site, could leave an imprint in the difference map even if nothing has bound. This is because instead of vacuum there might be bulk solvent occupying that space - and the positive Fo-Fc density will therefore show the missing bulk solvent contribution. Two possible suggestions for avoiding being mislead by this:
    • use a PDB file that describes the binding site but doesn't resemble the expected ligand. This could be a PDB file with randomly placed atoms within the region of interest or a PDB file of a significantly different compound (or a concatenation of several compounds binding in that site)
    • don't be tempted to look at the difference (Fo-Fc) maps at too low a contouring level: if nothing has bound the excluded bulk solvent density will start becoming visible at low contour levels. A better approach is to start at very high contour levels and lower them gradually.


While being aware of the limitations and potential biasing issue, the procedures described above can still be very useful. It is important to know what question one is actually asking:

  • "Has something bound and where?"
    • use the –L option
  • "Has something bound in site X?"
    • use first the –L option to confirm that something has bound (and hopefully in the known binding site)
    • use a generic site-describing PDB file with –Lpdb to potentially enhance the contrast in a second run. However, often this is not actually necessary, since the density is already good enough to model the ligand/compound in.

Once those questions have been answered, the ligand can be modelled into the binding site and refined normally.