Newsletter contents...

Datasets to Maps: A Wrapper for SHELXS and the CCP4 Program Suite

Paul Emsley

1. Introduction

The procedure of creating an initial map from MIR(AS) or MAD data is beset with traps for the novice crystallographer (and others not so novice). This program attempts to remove the drudge and pitfalls of heavy atom structure solution. I am aiming for a 90% success rate (i.e. Chart finds an acceptable map for at least 90% of soluble cases). Chart may create a final map that is worse than an experienced SHARP user could create. But it will create it much quicker.

Chart first uses the Patterson overlap method implemented in SHELXS to find initial heavy atom positions, running over various resolution ranges (typically more than one would do ``by hand''). Acceptable SHELXS solutions are then exhausively refined with MLPHARE. Atoms that refine well are combined. A cross-phased difference Fourier is synthesised and peaks checked to search for other derivatives. The solutions are then combined and the phases passed to DM. DM parameters are optimised for the best map. ⁵

One needs to specify:

• the Heavy Atom for the derivatives
• the number of residues in the protein
• the file-names for the post-truncate MTZ files
• and whether the structure solution should be MAD or MIR(AS).
  Chart seems to work for SIRAS data too.

2. Screenshot

3. Execution

As you would expect for a program that solves structures, Chart uses many programs of the CCP4 Program Suite. Most heavily used is MLPHARE.

Chart creates conventional csh ``com'' and ``log'' files for programs in the CCP4 suite, so that it makes it easy to see what is going on with the refinements and so that you can tweak the refinement after Chart has finished (e.g. by adding, for example, new derivative sites or reducing the solvent content). Chart may not infallibly get all parameters correct.

3.1 Implementation

Chart comes in two parts (at the time of writing). chart.scsh is code for the Scheme shell, scsh, and chart.stk is code for the GUI written for STk ⁶ (Scheme with Tk extensions). Hopefully soon I will be able to integrate the two by using guile, a Scheme interpreter with both scsh shell and Tk extensions. You will of course need both STk and scsh if you wish to use the interpreters, but binary distributions ⁷ for various sytems will be provided on the web site, for those wishing to avoid these compilation/installation steps.

The current implementation of chart runs CCP4 programs single-threadedly in the foreground. However, Chart can be changed to submit multiple jobs to remote processors ⁸. Chart is intensely parallelizable. Basically speed is proportional to the number of nodes or processors. I imagine that one could solve typical 50kD proteins in less than 5 minutes with 10 or so nodes i.e. processors).

3.2 Execution time

The execution time is dominated by the time taken to run the CCP4 programs. The time spent in the shell is trivial in comparison.

4. MAD

I have not tested the program with MAD data, although I hope to soon. Some changes will be needed of course, but largely the procedure remains unchanged. I need to include an interface to rantan and revise.

5. Future

• Add MAD analysis, rantan, revise.
• Integrate the Chart GUI into the main program
• Add dimer-searching [note: perhaps Kevin Cowtan has done this already with new DM?]
• Add map analysis program (score map according to ``the variance of variance of the map density'').
• Add skeletonisation map scoring.
• Test parallelization.
• Integrate into ccp4i
• Change to guile Scheme interpeter.

6. Where to get it

Chart is copyright Paul Emsley and is released under the GNU GPL. It will be available for free.

At the time of writing, Chart has not been released to the general public. It may have been by the time you read this however, but I don't want to release a first version that is very broken. See the web site for more details and updates:

http://www.chem.gla.ac.uk/~paule/chart

Newsletter contents...