Trying to cut the resolution of the datasets

From: Abd Ghani Abd Aziz
Date: 19 March 2012 13:59

hello everyone,

I am new in this bulletin board. I would like to know on how to cut my resolution in my datasets that have been processed/produced in diamond light source. In my processed directory, I found there are 3 files (free.mtz, scaled.sca and unmerged.sca). May I know which one can be used to cut my data that was diffracted to 1.5A? Cheers

regards

Abd Ghani

The University of Sheffield

----------
From: Eleanor Dodson


Qs

1) Why do you want to limit your data?

Most applications allow you to only use a specified sub-set - see GUI tasks for "resolution limits".

In general you may want to run moleculer replacement or exptl phasing at a limited resolution, but for refinenement or phase extension it is good to use the whole range..

Eleanor

--

----------
From: Graeme Winter


... presuming of course the automated software got this resolution limit right.

If for whatever reason you would like to cut the limit mtzutils will
do this nicely:

mtzutils hklin blah_free.mtz hklout blah_lower.mtz << eof
resolution 1.8
eof

(say) - I am sure there are other ways within the suite to do this.

Best wishes,

Graeme

----------
From: Tim Gruene 

Dear Abd Ghani,

The method described by Graeme is how the resolution can be delimited
artificially.
If you want to get the best from your data, determine the resolution
limit of your data e.g. with pointless (I/sigI > 2.0 is a good marker)
and reprocess the data to that limit. If you integrate the whole
detector area and the outer parts contain only noise, the noise has a
negative effect on the real data.

Tim

- --
- --

----------
From: Graeme Winter

Hi Tim,

That's interesting. When I looked at this (and I would say I looked
reasonably carefully) I found it only made a difference in the scaling
- integrating across the whole area was fine. However, I would expect
to see a difference, and likely an improvement, in scaling only the
data you want. It would also give you sensible merging statistics
which you'll probably want when you come to publish or deposit.

Abd Ghani: if you'd like to rerun the processing at Diamond to a
chosen resolution limit I will be happy to send some instructions.
That's probably a good idea.

Best wishes,

Graeme

----------
From: Pete Meyer


Hi Graeme,

That's interesting. When I looked at this (and I would say I looked
reasonably carefully) I found it only made a difference in the scaling
- integrating across the whole area was fine. However, I would expect
to see a difference, and likely an improvement, in scaling only the
data you want. It would also give you sensible merging statistics
which you'll probably want when you come to publish or deposit.

I've seen low-resolution datasets where the resolution cutoff had a fairly significant impact on integration - usually in cell or orientation refinement.  This seemed to make sense, as trying to use large numbers of spots that weren't really there (and so had essentially undetermined positions) tended to lead to instabilities in refinement.

What resolution ranges were you looking at?

Pete

----------
From: Graeme Winter


Hi Pete,

A couple of observations with this. I found when doing cell refinement
with Mosflm setting a conservative resolution limit *was* very helpful
in making sure that the refinement was stable. But then switching back
to the full detector area for integration was fine after that. However
- and this is a  big however, the data analysed for this were
typically reasonably high resolution i.e. 3 - 1.6 A.

I've not tried the same thing with a lot of lower resolution data -
there are not many people out there making loads of useful 6A data
sets public!

Best wishes,

Graeme