Re: [ccp4bb] Do my SAXS data agree with the crystal structure?
Dear Xun, Regarding your monomer vs dimer, theoretical vs observed crysol plots - yes - they are significantly different. If you focus at the very lowest q part of the curve - the deviation there in your monomer plots indicate that there is a significant size difference between your PX monomer and your SAXS data - the PX dimer is a much better fit at low q. This should be enough to demonstrate to a reviewer that the dimer you see in PX is also present in solution. Other experiments that could support this are SEC-MALLS or perhaps AUC. HTH, Dave David C. Briggs PhD Father, Structural Biologist and Sceptic University of Manchester E-mail: david.c.bri...@manchester.ac.uk Webs : http://flavors.me/xtaldave Twitter: @xtaldave Skype: DocDCB On 17 June 2012 06:11, Xun Lu xlun...@gmail.com wrote: Drs.Caldwell, Briggs, and Gupta, Thank you very much for the advices. I regret that I didn't show any figure in the earlier post. Here I've attached a figure showing the data quality and some fittings. Data look OK, right? This question may sound silly, but I just want to make sure. As I said in the earlier post, I tried Crysol. I used the crystal structure (dimer+DNA) as the model, and the fitting was OK, right? In fact, I also tried monomer+DNA as the model (I simply deleted one monomer from the PDB file). This kind of comparison may be meaningless, but I was just curious. I am wondering how people judge whether the fit is good or not. Another question, I tried to generate an envelope from SAXS data using Gasbor and Dammin (people say Dammin is better at protein-DNA complex, although it still uses the same bead for both DNA and protein?). The generated envelope was nothing like my crystal structure. As people have pointed out, protein and DNA scatter differently. SANS is the way to go. So I should give up on modeling SAXS data? I've almost given up, because anyways I have the crystal structure, and SAXS is only a small part of this paper. Thanks, Xun On Sat, Jun 16, 2012 at 6:36 PM, Kushol Gupta kushol.gu...@gmail.com wrote: Two cents - A good deal of caution must be exercised when working with composite particles such as a protein-DNA complex in SAXS because of the contrast problem. Simply, protein and DNA scatter differently in x-rays, with a bias towards the DNA component. As a result, experimental Rgs could be slightly deflated versus what their true values would be at infinite contrast. Mass estimation by I(0) analysis with a protein standard of known mass and concentration is not really valid because the contrast terms are different. Because the particle is heterogeneous in composition and distribution, shape reconstruction from SAXS alone, which assumes homogeneity, can also be misleading (although in practice it is still reasonably instructive). It is for these reasons that SANS and the contrast variation approach can be extremely useful. With those caveats, the strategy you describe - comparison of experimental and theoretical profiles from an experimental structure using CRYSOL or FoxS is definitely the best way to go in the case of a protein-DNA complex with SAXS alone. Showing comparisons of the experimental with the calculated should make the point. Test other possible models inferred from lattice packing to further your point (if applicable). Regarding populations of monomer and dimer - · it is generally good to constrain your interpretation of scattering data with other orthogonal solution measures which demonstrates the homogeneity of your complex in comparable experimental conditions, such as sedimentation velocity or gel filtration. · Have some determination of affinity of the complex in the same solution conditions (including temperature!). This will allow you to argue that your sample concentrations are well in excess of any monomer-dimer association behavior (eg, mixtures!). Scattering of mixtures can undermine your ability to accurately assess the structural properties of your complex. · Collect a concentration series and extrapolate to infinite dilution, if possible, to ensure elimination of the S(q) term from your data. Interparticle interactions can be an issue with complexes containing DNA if the buffers aren’t quite right. (I’ve seen this a lot) Lastly, remember that the scattering profile represents the solution average of the particle, not just a single snapshot. Some discrepancies like those you note should be expected. Hope that helps, Kushol Kushol Gupta, Ph.D. Research Associate - Van Duyne Laboratory HHMI / Perelman School of Medicine University of Pennsylvania kgu...@mail.med.upenn.edu 215-573-7260 / 267-259-0082 -Original Message- From: CCP4 bulletin board
Re: [ccp4bb] Do my SAXS data agree with the crystal structure?
If you main point is dimer vs monomer, Guinier plots is in fact all you would need, in my opinion according to what I have read so far... A. On 17 Jun 2012, at 13:01, David Briggs wrote: Dear Xun, Regarding your monomer vs dimer, theoretical vs observed crysol plots - yes - they are significantly different. If you focus at the very lowest q part of the curve - the deviation there in your monomer plots indicate that there is a significant size difference between your PX monomer and your SAXS data - the PX dimer is a much better fit at low q. This should be enough to demonstrate to a reviewer that the dimer you see in PX is also present in solution. Other experiments that could support this are SEC-MALLS or perhaps AUC. HTH, Dave David C. Briggs PhD Father, Structural Biologist and Sceptic University of Manchester E-mail: david.c.bri...@manchester.ac.uk Webs : http://flavors.me/xtaldave Twitter: @xtaldave Skype: DocDCB On 17 June 2012 06:11, Xun Lu xlun...@gmail.com wrote: Drs.Caldwell, Briggs, and Gupta, Thank you very much for the advices. I regret that I didn't show any figure in the earlier post. Here I've attached a figure showing the data quality and some fittings. Data look OK, right? This question may sound silly, but I just want to make sure. As I said in the earlier post, I tried Crysol. I used the crystal structure (dimer+DNA) as the model, and the fitting was OK, right? In fact, I also tried monomer+DNA as the model (I simply deleted one monomer from the PDB file). This kind of comparison may be meaningless, but I was just curious. I am wondering how people judge whether the fit is good or not. Another question, I tried to generate an envelope from SAXS data using Gasbor and Dammin (people say Dammin is better at protein-DNA complex, although it still uses the same bead for both DNA and protein?). The generated envelope was nothing like my crystal structure. As people have pointed out, protein and DNA scatter differently. SANS is the way to go. So I should give up on modeling SAXS data? I've almost given up, because anyways I have the crystal structure, and SAXS is only a small part of this paper. Thanks, Xun On Sat, Jun 16, 2012 at 6:36 PM, Kushol Gupta kushol.gu...@gmail.com wrote: Two cents - A good deal of caution must be exercised when working with composite particles such as a protein-DNA complex in SAXS because of the contrast problem. Simply, protein and DNA scatter differently in x-rays, with a bias towards the DNA component. As a result, experimental Rgs could be slightly deflated versus what their true values would be at infinite contrast. Mass estimation by I(0) analysis with a protein standard of known mass and concentration is not really valid because the contrast terms are different. Because the particle is heterogeneous in composition and distribution, shape reconstruction from SAXS alone, which assumes homogeneity, can also be misleading (although in practice it is still reasonably instructive). It is for these reasons that SANS and the contrast variation approach can be extremely useful. With those caveats, the strategy you describe - comparison of experimental and theoretical profiles from an experimental structure using CRYSOL or FoxS is definitely the best way to go in the case of a protein-DNA complex with SAXS alone. Showing comparisons of the experimental with the calculated should make the point. Test other possible models inferred from lattice packing to further your point (if applicable). Regarding populations of monomer and dimer - · it is generally good to constrain your interpretation of scattering data with other orthogonal solution measures which demonstrates the homogeneity of your complex in comparable experimental conditions, such as sedimentation velocity or gel filtration. · Have some determination of affinity of the complex in the same solution conditions (including temperature!). This will allow you to argue that your sample concentrations are well in excess of any monomer-dimer association behavior (eg, mixtures!). Scattering of mixtures can undermine your ability to accurately assess the structural properties of your complex. · Collect a concentration series and extrapolate to infinite dilution, if possible, to ensure elimination of the S(q) term from your data. Interparticle interactions can be an issue with complexes containing DNA if the buffers aren’t quite right. (I’ve seen this a lot) Lastly, remember that the scattering profile represents the solution average of the particle, not just a single snapshot. Some discrepancies like those you note should be expected. Hope that helps, Kushol Kushol
[ccp4bb] Fwd: [ccp4bb] Do my SAXS data agree with the crystal structure?
In theory, there should be a simple way to calculate P(r) directly from the crystal structure rather than indirectly from the expected scattering curve. Distribution of pair distances, r^2 weighted. This would remove any ambiguity about choice of Dmax. ... but I can't think of any of the common SAXS programs that do it that way. Clearly, since you have the crystal structure, you know the exact Dmax (the maximum diameter of the object). I would use the maximum atom pair distance from the crystal structure, then add a little bit to account for the width of the atoms and solvation layer. Do this only for the theoretical curve. The Dmax for the solution structure may be different due to conformational fluctuations. Dmax is not a well-defined quantity in reality and has a large error range. I would describe the feature you see in the theoretical P(r) at 70A as a shoulder. Such a feature in a dimer is not surprising because you have two large domains separated by a distance. Do you see it in the monomer P(r)? The fact that it is smoother in the solution data is also not surprising, since you can expect domains to move around on average. Some programs actually attempt to model this kind of disorder by sampling conformation space to see which various conformations best fit the curve. Probably that's more analysis than would be useful to you. Richard On Jun 17, 2012, at 1:11 AM, Xun Lu wrote: Drs.Caldwell, Briggs, and Gupta, Thank you very much for the advices. I regret that I didn't show any figure in the earlier post. Here I've attached a figure showing the data quality and some fittings. Data look OK, right? This question may sound silly, but I just want to make sure. As I said in the earlier post, I tried Crysol. I used the crystal structure (dimer+DNA) as the model, and the fitting was OK, right? In fact, I also tried monomer+DNA as the model (I simply deleted one monomer from the PDB file). This kind of comparison may be meaningless, but I was just curious. I am wondering how people judge whether the fit is good or not. Another question, I tried to generate an envelope from SAXS data using Gasbor and Dammin (people say Dammin is better at protein-DNA complex, although it still uses the same bead for both DNA and protein?). The generated envelope was nothing like my crystal structure. As people have pointed out, protein and DNA scatter differently. SANS is the way to go. So I should give up on modeling SAXS data? I've almost given up, because anyways I have the crystal structure, and SAXS is only a small part of this paper. Thanks, Xun
Re: [ccp4bb] Do my SAXS data agree with the crystal structure?
It might also be useful to go through the recent paper by Jacques et al Acta cryst D68:620-6 (2012) towards the standardization of saxs data analysis, diagnostics and reporting. Best regards, Savvas On 17-jun.-2012, at 13:01, David Briggs drdavidcbri...@gmail.com wrote: Dear Xun, Regarding your monomer vs dimer, theoretical vs observed crysol plots - yes - they are significantly different. If you focus at the very lowest q part of the curve - the deviation there in your monomer plots indicate that there is a significant size difference between your PX monomer and your SAXS data - the PX dimer is a much better fit at low q. This should be enough to demonstrate to a reviewer that the dimer you see in PX is also present in solution. Other experiments that could support this are SEC-MALLS or perhaps AUC. HTH, Dave David C. Briggs PhD Father, Structural Biologist and Sceptic University of Manchester E-mail: david.c.bri...@manchester.ac.uk Webs : http://flavors.me/xtaldave Twitter: @xtaldave Skype: DocDCB On 17 June 2012 06:11, Xun Lu xlun...@gmail.com wrote: Drs.Caldwell, Briggs, and Gupta, Thank you very much for the advices. I regret that I didn't show any figure in the earlier post. Here I've attached a figure showing the data quality and some fittings. Data look OK, right? This question may sound silly, but I just want to make sure. As I said in the earlier post, I tried Crysol. I used the crystal structure (dimer+DNA) as the model, and the fitting was OK, right? In fact, I also tried monomer+DNA as the model (I simply deleted one monomer from the PDB file). This kind of comparison may be meaningless, but I was just curious. I am wondering how people judge whether the fit is good or not. Another question, I tried to generate an envelope from SAXS data using Gasbor and Dammin (people say Dammin is better at protein-DNA complex, although it still uses the same bead for both DNA and protein?). The generated envelope was nothing like my crystal structure. As people have pointed out, protein and DNA scatter differently. SANS is the way to go. So I should give up on modeling SAXS data? I've almost given up, because anyways I have the crystal structure, and SAXS is only a small part of this paper. Thanks, Xun On Sat, Jun 16, 2012 at 6:36 PM, Kushol Gupta kushol.gu...@gmail.com wrote: Two cents - A good deal of caution must be exercised when working with composite particles such as a protein-DNA complex in SAXS because of the contrast problem. Simply, protein and DNA scatter differently in x-rays, with a bias towards the DNA component. As a result, experimental Rgs could be slightly deflated versus what their true values would be at infinite contrast. Mass estimation by I(0) analysis with a protein standard of known mass and concentration is not really valid because the contrast terms are different. Because the particle is heterogeneous in composition and distribution, shape reconstruction from SAXS alone, which assumes homogeneity, can also be misleading (although in practice it is still reasonably instructive). It is for these reasons that SANS and the contrast variation approach can be extremely useful. With those caveats, the strategy you describe - comparison of experimental and theoretical profiles from an experimental structure using CRYSOL or FoxS is definitely the best way to go in the case of a protein-DNA complex with SAXS alone. Showing comparisons of the experimental with the calculated should make the point. Test other possible models inferred from lattice packing to further your point (if applicable). Regarding populations of monomer and dimer - · it is generally good to constrain your interpretation of scattering data with other orthogonal solution measures which demonstrates the homogeneity of your complex in comparable experimental conditions, such as sedimentation velocity or gel filtration. · Have some determination of affinity of the complex in the same solution conditions (including temperature!). This will allow you to argue that your sample concentrations are well in excess of any monomer-dimer association behavior (eg, mixtures!). Scattering of mixtures can undermine your ability to accurately assess the structural properties of your complex. · Collect a concentration series and extrapolate to infinite dilution, if possible, to ensure elimination of the S(q) term from your data. Interparticle interactions can be an issue with complexes containing DNA if the buffers aren’t quite right. (I’ve seen this a lot) Lastly, remember that the scattering profile represents the solution average of the particle, not just a single snapshot. Some discrepancies like
