DM for skeletonisation (CCP4: General)

NAME

dm_skeletonisation - iterative skeletonisation using dm

This document describes the iterative skeletonisation process as implemented in 'dm', and in particular the things you need to know in order to use the process.

THE SKELETONISATION PROCESS

The process of skeletonisation involves two stages:
  1. The construction of a 'skeleton' representing the connectivity of the map.
  2. The creation of a new map from the skeleton.

These stages are conducted as follows:

1. Construction of the Skeleton

First, the map is searched to locate the important features. These fall into two classes: Peaks, and Join points.

Peaks are local maxima in the density. A ridge is the highest path joining two peaks. A join point is the lowest point on a ridge, i.e. the lowest point on the highest path joining two peaks.

In constructing the skeleton, there are two important parameters to consider:

  1. The join point cutoff. Any ridge whose join point falls above this level will be included in the skeleton. Ridges whose join points fall beneath this level will not be included in full, but a part of the ridge may be included as an end.
  2. The end point cutoff. If a join point is below the join point cutoff, a portion of the ridge may still be added to the skeleton if the peak is above the end point cutoff. That portion of the ridge which is above this level is added to the skeleton.

Obviously, ridges included under (1) will contribute to the connectivity of the map, whereas ridges included under (2) will not because they are always dead-ends. Thus it is normal to set the end point cutoff somewhat higher than the join point cutoff in order to enhance connectivity in the map.

To simplify the setting of these parameters, 'dm' determines the cutoff levels by examining the map in order to produce a skeleton with particular features. These features are specified as the length of skeleton (per residue) which should be generated from join point ridges, and the length of skeleton (per residue) which should be generated from dead-end ridges. It is my hope that these optimum values of these parameters should be fairly constant, and so that for most cases the default values should suffice. If new values are required, they can be set using the SKEL LENGTH card.

One other parameter affects the skeletonisation process: the temperature factor applied to the initial map. Skeletonisation based on the sharpened map used by 'dm' is fairly poor, a much better skeleton is obtained if the map is smoothed with an overall temperature factor first. Again this parameter is not very sensitive and can be left at the default value, you may possibly want to increase it if the initial map is particularly poor or decrease it if the initial map is very good.

2. Map Generation

Map generation is carried out by building up a cylindrically symmetric density distribution about the skeleton. The distribution is taken from an atom averaged over a cylinder whose length is the average protein atomic bond length.

Density outside the solvent mask is set to the mean solvent value.

USING SKELETONISATION

In skeletonisation calculations I have found the best scheme is to perform a conventional density modification first, and then use the modified map as a starting point for a dm/skeletonisation calculation. The phase extension scheme 'SCHEME AUTO FRAC 0.5' is a good option for this second calculation, as reflections with high figures-of-merit are more likely to be included in the starting set.

Note also that it is wise to make sure the final cycle of the skeletonisation run is a conventional density modification cycle rather than a skeletonisation cycle otherwise your building may become excessively biased towards the 'dm' skeleton. Thus for 'SKEL EVERY 2' you would set 'NCYC' to be odd, for 'SKEL EVERY 3' (the default) you would set 'NCYC' to be (3n-1) where n is an integer.

A typical command file is given below:


dm     
    hklin ~/hkl/gmtomir.mtz    
    hklout $SCRATCH/gmt_dm.mtz 
  << 'MY-DATA'
SOLC 0.35
MODE SOLV HIST
NCYC 10
SCHEME RES FROM 3.7
LABI FP=FP SIGFP=SIGFP PHIO=PHIB FOMO=FOM
LABO PHIDM=PHIdm FOMDM=FOMdm
END
'MY-DATA'
#
dm     
    hklin $SCRATCH/gmt_dm.mtz  
    hklout $SCRATCH/gmt_sk.mtz 
  << 'MY-DATA'
SOLC 0.35
MODE SOLV HIST SKEL
SKEL EVERY 2
NCYC 9
SCHEME AUTO FRAC 0.5
LABI FP=FP SIGFP=SIGFP PHIO=PHIB FOMO=FOM  PHIDM=PHIdm FOMDM=FOMdm
LABO PHIDM=PHIsk FOMDM=FOMsk
END
'MY-DATA'

AUTHOR

Dr. Kevin Cowtan, Protein Structure Group, University of York, England