[ccp4bb] Anisotropic data and an extremely long c axis
Hi, I also have a question concerning anisotropic data. Collected a data set and the best crystal gave highly anisotropic diffraction patterns ( 3.7 A - 5.8 A). So my first question is how to handle these data. I got only experience with normal data using the ccp4 suite. Are there any program specially for these kind of data? There are? The second question is how anisotropic data occur? The protein I work with has a tetragonal sg with a=b= 86.0 and an extremely long c axis of 651 A. Secondary Structure prediction suggest a lot of beta strands. How can I explain the anisotropy (for my own interest and my thesis)? Thank you very much. Marie
Re: [ccp4bb] Anisotropic data and an extremely long c axis
A first concern with that extreme anisotropy is at the integration and scaling stages. Large swaths of your detector are empty of reflections, but they will still bias the way reference profiles are calculated at integration; while the lots of reflections with intensities around 0 (but with significant sigmas) will bias the statistics for scaling the real reflections (real as in having intensity). You are better off excluding all the parts of the detector that do not contain significant intensities (and taking a hit in completeness) than trying to correct for all the 0 intensities. While the optimum would be to limit integration to an elliptical area of the detector containing the significant intensities, I am not aware of any program that can do this elegantly. In my experience, the strong data almost always end in the horizontal direction of the detector. And in many cases, the significant area while really elliptical, can be approximated to a rectangle, with the long edges horizontal. In that case, a simple trick is to use the excluded areas in the integration programs to limit integration to your rectangle. In Denzo I usually did this by playing with the detector distance, so that the diffraction ends at the edges of the detector, effectively setting the 2 short sides of our rectangle, and set the long edges of the rectangle with the excluded rectangle and excluded circle options (a circle of very large radius centered way out of the detector can be an excellent approximation to a straight line for cutting purposes). I am sure most programs will allow for equivalent ways of limiting integration to a rectangle. Once you have an mtz file with intensities, the UCLA anisotropy correction server can be a great help too. Marie Lacroix lacroix.ma...@rocketmail.com wrote: Hi, I also have a question concerning anisotropic data. Collected a data set and the best crystal gave highly anisotropic diffraction patterns ( 3.7 A - 5.8 A). So my first question is how to handle these data. I got only experience with normal data using the ccp4 suite. Are there any program specially for these kind of data? There are? The second question is how anisotropic data occur? The protein I work with has a tetragonal sg with a=b= 86.0 and an extremely long c axis of 651 A. Secondary Structure prediction suggest a lot of beta strands. How can I explain the anisotropy (for my own interest and my thesis)? Thank you very much. Marie -- *** Jose Antonio Cuesta-Seijo Biophysical Chemistry Group Department of Chemistry University of Copenhagen Tlf: +45-35320261 Universitetsparken 5 DK-2100 Copenhagen, Denmark ***
Re: [ccp4bb] Anisotropic data and an extremely long c axis
Anisotropy in the diffraction pattern could simply be due to the shape of the crystals. The intensity of diffraction is a function of the volume of diffracting matter that is hit by the X-ray beam. Think for example of a thin plate crystal, which you rotate in the X-ray beam. When the plate is perpendicular to the X-ray beam, the volume of matter hit by the X-rays is much smaller than when the plate is parallel to the X-ray beam. When processing data using XDS, at the integration level (INTEGRATE) there is for each frame a single scale factor that is reported (I cannot tell you what the INTEGRATE.LP file says exactly - I am sitting at a conference where I cannot access my data files at home), but you can follow the increase/decrease in intensity of diffraction (and thus the volume of diffracting matter in the X-ray beam) by following the variation of these scale factors. Otherwise, sometimes (once the structure is solved) it is possible, 'a posteriori', to give a plausible explanation to such an anisotropy. For example, by noticing anisotropic crystal contacts (e.g. multiple contacts along two directions, and very few crystal forming contacts with plenty of solvent in the third direction). HTH, Fred. Message du 09/06/10 14:33 De : Marie Lacroix A : CCP4BB@JISCMAIL.AC.UK Copie à : Objet : [ccp4bb] Anisotropic data and an extremely long c axis Hi, I also have a question concerning anisotropic data. Collected a data set and the best crystal gave highly anisotropic diffraction patterns ( 3.7 A - 5.8 A). So my first question is how to handle these data. I got only experience with normal data using the ccp4 suite. Are there any program specially for these kind of data? There are? The second question is how anisotropic data occur? The protein I work with has a tetragonal sg with a=b= 86.0 and an extremely long c axis of 651 A. Secondary Structure prediction suggest a lot of beta strands. How can I explain the anisotropy (for my own interest and my thesis)? Thank you very much. Marie
Re: [ccp4bb] Anisotropic data and an extremely long c axis
Hi Many years ago I coded up integration using anisotropic resolution limits for Mosflm - it seemed to work well, but the refinement programs available at the time really didn't like huge regions of reciprocal space having no data in them - they preferred to have measurements there with sigmas, even if the measurements were just background. I thought my implementation was rather elegant, since it integrated a rather nicely formed ellipsoidal region of reciprocal space. So although the code is still there, and it still works, I don't make a big deal about it. If the refinement programs are happy to deal with the unmeasured data (in the directions where the crystal doesn't diffract so well), I'm happy to put the effort in to resurrect it. As for how the anisotropy occurs, there are a few good reasons; as Fred said, the illuminated volume of the crystal can contribute. I think another point is that there is no reason why (for a non-cubic crystal) the order in the crystal should be isotropic; for example, if you have molecules that are approximate prolate spheroids (think rugby ball, or football for our American readers), they can obviously pack better with their long axes aligned, but the orientation about that long axis can be rather less well defined. The diffraction is a reflection (ahem) of the internal order... Hi, I also have a question concerning anisotropic data. Collected a data set and the best crystal gave highly anisotropic diffraction patterns ( 3.7 A - 5.8 A). So my first question is how to handle these data. I got only experience with normal data using the ccp4 suite. Are there any program specially for these kind of data? There are? The second question is how anisotropic data occur? The protein I work with has a tetragonal sg with a=b= 86.0 and an extremely long c axis of 651 A. Secondary Structure prediction suggest a lot of beta strands. How can I explain the anisotropy (for my own interest and my thesis)? Thank you very much. Marie Harry -- Dr Harry Powell, MRC Laboratory of Molecular Biology, Hills Road, Cambridge, CB2 0QH
Re: [ccp4bb] Anisotropic data and an extremely long c axis
Dear Marie, I believe that the first of Fred's explanations can mostly be corrected for by scaling (and it could partly be overcome by longer exposure times as long as radiation damage does not kick in). In your case, where one cell axis is about 10x as long as the other two, Fred's second explanation is probably the real cause of anisotropy: the unit cells would have to be in proper order in the c-direction over 10x the range compared to a/b in order to reach the same resolution. Tim On Wed, Jun 09, 2010 at 12:22:59PM +, Marie Lacroix wrote: Hi, I also have a question concerning anisotropic data. Collected a data set and the best crystal gave highly anisotropic diffraction patterns ( 3.7 A - 5.8 A). So my first question is how to handle these data. I got only experience with normal data using the ccp4 suite. Are there any program specially for these kind of data? There are? The second question is how anisotropic data occur? The protein I work with has a tetragonal sg with a=b= 86.0 and an extremely long c axis of 651 A. Secondary Structure prediction suggest a lot of beta strands. How can I explain the anisotropy (for my own interest and my thesis)? Thank you very much. Marie -- -- Tim Gruene Institut fuer anorganische Chemie Tammannstr. 4 D-37077 Goettingen GPG Key ID = A46BEE1A signature.asc Description: Digital signature
Re: [ccp4bb] AW: AW: AW: [ccp4bb] Anisotropic data and an extremely long c axis
I guess then Fred's reply part one explains everything :-) Jürgen - Jürgen Bosch Johns Hopkins Bloomberg School of Public Health Department of Biochemistry Molecular Biology Johns Hopkins Malaria Research Institute 615 North Wolfe Street, W8708 Baltimore, MD 21205 Phone: +1-410-614-4742 Lab: +1-410-614-4894 Fax: +1-410-955-3655 http://web.mac.com/bosch_lab/http://web.me.com/bosch_lab/ On Jun 9, 2010, at 10:48 AM, Marie Lacroix wrote: Yes, I just checked this. Von: Bosch, Juergen jubo...@jhsph.edumailto:jubo...@jhsph.edu An: Marie Lacroix lacroix.ma...@rocketmail.commailto:lacroix.ma...@rocketmail.com Gesendet: Mittwoch, den 9. Juni 2010, 16:01:56 Uhr Betreff: Re: AW: AW: [ccp4bb] Anisotropic data and an extremely long c axis Does this coincide with the direction of better diffraction ? Jürgen - Jürgen Bosch Johns Hopkins Bloomberg School of Public Health Department of Biochemistry Molecular Biology Johns Hopkins Malaria Research Institute 615 North Wolfe Street, W8708 Baltimore, MD 21205 Phone: +1-410-614-4742 Lab: +1-410-614-4894 Fax: +1-410-955-3655 http://web.mac.com/bosch_lab/http://web.me.com/bosch_lab/ On Jun 9, 2010, at 9:38 AM, Marie Lacroix wrote: The crystal dimensions are 275 x 150 x 100 µm. So in one dimension they are bigger. Von: Bosch, Juergen jubo...@jhsph.edumailto:jubo...@jhsph.edu An: Marie Lacroix lacroix.ma...@rocketmail.commailto:lacroix.ma...@rocketmail.com Gesendet: Mittwoch, den 9. Juni 2010, 15:24:51 Uhr Betreff: Re: AW: [ccp4bb] Anisotropic data and an extremely long c axis but thicker in one dimension ? Jürgen - Jürgen Bosch Johns Hopkins Bloomberg School of Public Health Department of Biochemistry Molecular Biology Johns Hopkins Malaria Research Institute 615 North Wolfe Street, W8708 Baltimore, MD 21205 Phone: +1-410-614-4742 Lab: +1-410-614-4894 Fax: +1-410-955-3655 http://web.mac.com/bosch_lab/http://web.me.com/bosch_lab/ On Jun 9, 2010, at 8:41 AM, Marie Lacroix wrote: No, they are really large and beautiful 3D crystals. Von: Jürgen Bosch jubo...@jhsph.edumailto:jubo...@jhsph.edu An: Marie Lacroix lacroix.ma...@rocketmail.commailto:lacroix.ma...@rocketmail.com Gesendet: Mittwoch, den 9. Juni 2010, 14:38:37 Uhr Betreff: Re: [ccp4bb] Anisotropic data and an extremely long c axis Flat crystals ? Is the high resolution diffraction because you are shooting through more crystal volume ? Just a thought Jürgen .. Jürgen Bosch Johns Hopkins Bloomberg School of Public Health Department of Biochemistry Molecular Biology Johns Hopkins Malaria Research Institute 615 North Wolfe Street, W8708 Baltimore, MD 21205 Phone: +1-410-614-4742 Lab: +1-410-614-4894 Fax: +1-410-955-3655 http://web.mac.com/bosch_lab/ On Jun 9, 2010, at 8:22, Marie Lacroix lacroix.ma...@rocketmail.commailto:lacroix.ma...@rocketmail.com wrote: Hi, I also have a question concerning anisotropic data. Collected a data set and the best crystal gave highly anisotropic diffraction patterns ( 3.7 A - 5.8 A). So my first question is how to handle these data. I got only experience with normal data using the ccp4 suite. Are there any program specially for these kind of data? There are? The second question is how anisotropic data occur? The protein I work with has a tetragonal sg with a=b= 86.0 and an extremely long c axis of 651 A. Secondary Structure prediction suggest a lot of beta strands. How can I explain the anisotropy (for my own interest and my thesis)? Thank you very much. Marie
Re: [ccp4bb] Anisotropic data and an extremely long c axis
Frederic VELLIEUX wrote: Anisotropy in the diffraction pattern could simply be due to the shape of the crystals. The intensity of diffraction is a function of the volume of diffracting matter that is hit by the X-ray beam. Think for example of a thin plate crystal, which you rotate in the X-ray beam. When the plate is perpendicular to the X-ray beam, the volume of matter hit by the X-rays is much smaller than when the plate is parallel to the X-ray beam. Fred, I think this is a bit misleading. Although diffraction anisotropy is often accompanied by platy or needle-y crystals, the crystal shape is neither necessary nor sufficient to produce an anisotropic B factor. You will get an anisotropic SCALE factor if bits of the crystal are moving in and out of the beam as you describe, but scale factors and B factors are not the same thing! B factors arise from differences between neighboring unit cells (within a few microns of each other), and anisotropic B factors arise when the average displacement of atoms from their ideal lattice points is higher in one direction than another. Admittedly, both scale and B can have an effect on the resolution limit, but the latter kills high-angle spots much more rapidly than the former. Operationally, I recommend treating anisotropic data just like isotropic data. There is nothing wrong with measuring a lot of zeros (think about systematic absences), other than making irrelevant statistics like Rmerge higher. One need only glance at the formula for any R factor to see that it is undefined when the true F is zero. Unfortunately, there are still a lot of reviewers out there who were trained that the Rmerge in the outermost resolution bin must be 20%, and so some very sophisticated ellipsoidal cut-off programs have been written to try and meet this criterion without throwing away good data. I am actually not sure where this idea came from, but I challenge anyone to come up with a sound statistical basis for it. Better to use I/sigma(I) as a guide, as it really does tell you how much information vs noise you have at a given resolution. -James Holton MAD Scientist
Re: [ccp4bb] Anisotropic data and an extremely long c axis
I vaguely recall an email from Kay Diderich about 3 years ago to this board but I couldn't find it, describing a neat method of distorting the diffraction image to meet the ellipsoidal characteristics of the anisotropic diffraction. But I might be confusion myself, anyhow Kay can you comment on this ? I think it involved a feature in Sharp which I never used. Jürgen - Jürgen Bosch Johns Hopkins Bloomberg School of Public Health Department of Biochemistry Molecular Biology Johns Hopkins Malaria Research Institute 615 North Wolfe Street, W8708 Baltimore, MD 21205 Phone: +1-410-614-4742 Lab: +1-410-614-4894 Fax: +1-410-955-3655 http://web.mac.com/bosch_lab/http://web.me.com/bosch_lab/ On Jun 9, 2010, at 11:49 AM, James Holton wrote: Frederic VELLIEUX wrote: Anisotropy in the diffraction pattern could simply be due to the shape of the crystals. The intensity of diffraction is a function of the volume of diffracting matter that is hit by the X-ray beam. Think for example of a thin plate crystal, which you rotate in the X-ray beam. When the plate is perpendicular to the X-ray beam, the volume of matter hit by the X-rays is much smaller than when the plate is parallel to the X-ray beam. Fred, I think this is a bit misleading. Although diffraction anisotropy is often accompanied by platy or needle-y crystals, the crystal shape is neither necessary nor sufficient to produce an anisotropic B factor. You will get an anisotropic SCALE factor if bits of the crystal are moving in and out of the beam as you describe, but scale factors and B factors are not the same thing! B factors arise from differences between neighboring unit cells (within a few microns of each other), and anisotropic B factors arise when the average displacement of atoms from their ideal lattice points is higher in one direction than another. Admittedly, both scale and B can have an effect on the resolution limit, but the latter kills high-angle spots much more rapidly than the former. Operationally, I recommend treating anisotropic data just like isotropic data. There is nothing wrong with measuring a lot of zeros (think about systematic absences), other than making irrelevant statistics like Rmerge higher. One need only glance at the formula for any R factor to see that it is undefined when the true F is zero. Unfortunately, there are still a lot of reviewers out there who were trained that the Rmerge in the outermost resolution bin must be 20%, and so some very sophisticated ellipsoidal cut-off programs have been written to try and meet this criterion without throwing away good data. I am actually not sure where this idea came from, but I challenge anyone to come up with a sound statistical basis for it. Better to use I/sigma(I) as a guide, as it really does tell you how much information vs noise you have at a given resolution. -James Holton MAD Scientist
Re: [ccp4bb] Anisotropic data and an extremely long c axis
On 09/06/2010 16:49, James Holton wrote: Operationally, I recommend treating anisotropic data just like isotropic data. There is nothing wrong with measuring a lot of zeros (think about systematic absences), other than making irrelevant statistics like Rmerge higher. One need only glance at the formula for any R factor to see that it is undefined when the true F is zero. Unfortunately, there are still a lot of reviewers out there who were trained that the Rmerge in the outermost resolution bin must be 20%, and so some very sophisticated ellipsoidal cut-off programs have been written to try and meet this criterion without throwing away good data. I am actually not sure where this idea came from, but I challenge anyone to come up with a sound statistical basis for it. Better to use I/sigma(I) as a guide, as it really does tell you how much information vs noise you have at a given resolution. So, if my outer shell has 10% reflections I/sigI10, 90% reflections I/sigI=1, will Mean(I/sigI) for that shell tend to 10 or 1? Presumably I'm calculating it wrong in my simulation (very naive: took average of all individual I/sigI), because for me it tends to 1. But if I did get it right, then how does Mean(I/sigI) tell me that 10% of my observations have good signal? phx.
Re: [ccp4bb] Anisotropic data and an extremely long c axis
Frank von Delft wrote: On 09/06/2010 16:49, James Holton wrote: Operationally, I recommend treating anisotropic data just like isotropic data. There is nothing wrong with measuring a lot of zeros (think about systematic absences), other than making irrelevant statistics like Rmerge higher. One need only glance at the formula for any R factor to see that it is undefined when the true F is zero. Unfortunately, there are still a lot of reviewers out there who were trained that the Rmerge in the outermost resolution bin must be 20%, and so some very sophisticated ellipsoidal cut-off programs have been written to try and meet this criterion without throwing away good data. I am actually not sure where this idea came from, but I challenge anyone to come up with a sound statistical basis for it. Better to use I/sigma(I) as a guide, as it really does tell you how much information vs noise you have at a given resolution. So, if my outer shell has 10% reflections I/sigI10, 90% reflections I/sigI=1, will Mean(I/sigI) for that shell tend to 10 or 1? Presumably I'm calculating it wrong in my simulation (very naive: took average of all individual I/sigI), because for me it tends to 1. But if I did get it right, then how does Mean(I/sigI) tell me that 10% of my observations have good signal? It doesn't. The mean will not tell you anything about the distribution of I/sigI values, it will just tell you the average. If I may simplify your example case to: one good observation (I/sigI = 10) and 9 weak observations (I/sigI = 1), then Mean(I/sigI) = ~2. This is better than Mean(I/sigI) = 1, but admittedly still not great. I know it is tempting to say: but wait! I've got one really good reflection at that resolution! Doesn't that count for something? Well, it does (a little), but one good reflection does not a clear map make. -James Holton MAD Scientist
Re: [ccp4bb] Anisotropic data and an extremely long c axis
But at some point, getting a clear map might not be the goal. If you're in refinement mode, the weak reflections also provide information that your model needs to fit. I find I/sig(I) (or I/sig(I)) to be about as useful as Rmerge (or its relatives). Ron On Wed, 9 Jun 2010, James Holton wrote: Frank von Delft wrote: On 09/06/2010 16:49, James Holton wrote: Operationally, I recommend treating anisotropic data just like isotropic data. There is nothing wrong with measuring a lot of zeros (think about systematic absences), other than making irrelevant statistics like Rmerge higher. One need only glance at the formula for any R factor to see that it is undefined when the true F is zero. Unfortunately, there are still a lot of reviewers out there who were trained that the Rmerge in the outermost resolution bin must be 20%, and so some very sophisticated ellipsoidal cut-off programs have been written to try and meet this criterion without throwing away good data. I am actually not sure where this idea came from, but I challenge anyone to come up with a sound statistical basis for it. Better to use I/sigma(I) as a guide, as it really does tell you how much information vs noise you have at a given resolution. So, if my outer shell has 10% reflections I/sigI10, 90% reflections I/sigI=1, will Mean(I/sigI) for that shell tend to 10 or 1? Presumably I'm calculating it wrong in my simulation (very naive: took average of all individual I/sigI), because for me it tends to 1. But if I did get it right, then how does Mean(I/sigI) tell me that 10% of my observations have good signal? It doesn't. The mean will not tell you anything about the distribution of I/sigI values, it will just tell you the average. If I may simplify your example case to: one good observation (I/sigI = 10) and 9 weak observations (I/sigI = 1), then Mean(I/sigI) = ~2. This is better than Mean(I/sigI) = 1, but admittedly still not great. I know it is tempting to say: but wait! I've got one really good reflection at that resolution! Doesn't that count for something? Well, it does (a little), but one good reflection does not a clear map make. -James Holton MAD Scientist
