[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [ccp4bb]: problem to find the MR solution
*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
Dear David and others,
Thank you very much for your suggestions.
My protein does contain a large domain and a samll domain (about 20kDa). There are relative movements between the two domains in different complexes. I had tried using either the whole protein or the large domain as the searching model. Also, I tried to use the whole model, the polyalanine, or the only C-alpha as the starting model. But neither helped.
In terms of the space group, I spent a lot of times to check other possibilities. Auto index from Denzo clearly indicated it belongs to I-centered tetragonal. But, of course, it could also be indexed in orthorhombic and monoclinic. Though the index can be done in space group I422, I222, F222, and C2, the scaling gave much higher R-merge factors (>40% for both the overall R-factor and the individual linear R-factors), presumably due to the lack of two-fold axes along b- and c-directions.
For space group I4(1), the systematic absence condition requires that for reflection 00L: L=4N only. In my data, I have more than ten 00L reflections (over 10 I/sigma) that do not meet this condition. So, it seems clear that the space group should be I4, not I4(1).
With AMore, I played around a lot of parameters, such as the box and sphere sizes, resolution limit, etc. However, the correlation coefficients of the solutions (around 0.3-0.35) appeared too low to me. I also did rigidbody refinement for the top peaks from the translation searches, the cor. coef. increased to about 48%, but the r-factor only decreased to about 53%. There was no peak that was superior than others.
The BEAST program may be a good shot to try. I will try that when the next release of CCP4 is coming.
Thank you all for your help.
Jianping
----- Original Message -----
From: "David J. Schuller" <djs63@cornell.edu>
To: "Jianping Ding" <ding@cabm.rutgers.edu>
Cc: "ccp4 admin" <CCP4@daresbury.ac.uk>
Sent: Wednesday, February 27, 2002 12:21 AM
Subject: Re: [ccp4bb]: problem to find the MR solution
>
> RE: Jianping Ding's attempt to solve 90 kDa structure with MR
>
> getting the space group right is important, make sure you've done that
> before putting too much effort into later steps.
>
> 90 kDa is rather large. is the protein one big blob, or are there obvious
> divisions into domains? if so, you could search separately for the
> individual domains.
>
> using AMore:
> i) did you choose box and sphere sizes appropriate for such a
> large molecule? read the documentation for details, and check out CCP4
> script amore-sphere.
> ii) with AMoRe the correct rotation solution doesn't have to be at the top
> of the list, it just has to be on the list. if you used appropriate
> parameters don't sweat this and move onto the translation.
> iii) you could try running each of your translation
> function solutions through rigid body refinement. if you have a solution
> which is basically correct but is off by a degree or an Angstrom or
> so, this should improve the score substantially and make it stand out from
> background.
>
>
> Randy Read is working on a MR program named BEAST which is better than
> anything else available. it uses maximum likelihood scoring. i think it
> will be in the next version of CCP4, or contact him directly.
>
> cheers,
>
> =======================================================================
> "Supernaturalism is not, on the whole, friendly to science and technology,
> and there is no use in pretending that it is." - L. Sprague de Camp
> =======================================================================
> David J. Schuller
> modern man in a post-modern world
> MacCHESS, Cornell University
> djs63@cornell.edu
>
> On Tue, 26 Feb 2002, Jianping Ding wrote:
>
> > This is only partially related to CCP4, but I hope my fellow crystallographers subscribed to CCP4 list-server would like to give me help and advices.
> >
> > I have a protein of the size of about 90 kDa. I have been able to grow very tiny crystals from this protein (about 0.1x0.1x0.05 mm3 in dimension). The crystals diffracted X-ray better than 3 A. Even though the diffraction spots are pretty weak, the scaling of the data was pretty good to 3 A with the r-merge of 0.12 and completeness of 99%. The space group of the crystals is I4. Analysis of the systematic absence reflections indicated that there is no screw axis, i.e., the space group cannot be I4(1).
> >
> > I have a structure model from the same kind protein from a different species that was refined to 2.0 A. The sequence alignment of the two proteins indicated that all important structure elements are conserved. They share 44% identity in sequence with very few deletions or insertions (3%). Thus, I thought it would be easy and straightforward to solve the new structure using the MR method.
> >
> > However, all of my efforts using either CNS or AMORE programs to find the structure solution have failed so far.
> >
> > Using CNS, the rotation function search yielded a list of solutions in which the top peaks (over ten peaks) always have the RF-function value of 1.000 (I don¡¯t know what it means here). The subsequent translation search yielded a list of results with very poor monitor values and packing value. For space group I4, the translation search should carry out only along x- and y-axes. But CNS also does the translation search along z-axis.
> >
> > Using AMORE, both rotation function search and translation function search resulted no outstanding peaks. Most of the peaks in the list had correlation coefficient around 0.3-0.4 and R-factor around 0.55.
> >
> > I am wondering if there is something wrong with my data or there are other problems too. Does anyone in the community have similar experience? Does anyone have any clue on what¡¯s going wrong here? Or does anyone have any suggestions and advices for me?
>
>
>