Re: [ccp4bb] phenix.real_space_correlation vs overlapmap
Hi Matt,
What I want is the density 'inside' each atom in the map, so that I can make
> quantitative comparisons of "how much stuff" is present in a blob (modeled
> or not).
>
that's still ambiguous...: 1) do you want a density value calculated at a
point, which may be, for example, a center for an atom, or/and 2) you want
an array of points corresponding to the density values inside of a sphere
surrounding an atom of radius R (define R then - that's will be my next
question)?
> Are the per-residue "Fo" and "Fc" columns in the output of
>>> phenix.real_space_correlation actually what I am looking for?
>>
>>
>> I'm not sure what you mean... Could you please explain? If you want to
>> get electron density value (in sigma or e/A**3) computed at each atomic
>> center (or any specified point in space), then phenix.map_value_at_point
>> would do it.
>>
>
> Well the output of phenix.real_space_correlation has 'Fc' and 'Fo' columns
> like so:
>
> i_seq : PDB_string element occ b CC Fc Fo No.Points
> FLAG
> 0 : " N SER C1002 " N 1.00 28.79 0.9871 2.58 2.75 52
> 1 : " CA SER C1002 " C 1.00 29.56 0.9945 2.67 2.58 58
> 2 : " C SER C1002 " C 1.00 31.68 0.9966 3.28 3.22 56
> 3 : " O SER C1002 " O 1.00 28.79 0.9971 3.51 3.55 55
> 4 : " CB SER C1002 " C 1.00 36.38 0.9531 2.06 1.80 56
>
Ok, sound like you asked it to calculate the map CC given two maps, {Fc,
Phc} and {Fo, Phc}. The values reported in the above columns are the
eight-point interpolated density values of corresponding maps computed at
atomic centers of listed atoms.
> Does the 'Fc' column mean that 2.58 is the average density in atom number
> 0 (based on F_calc)?
>
No, it is the eight-point interpolated density value of Fc map computed at
atomic center of atom number 0.
> If so is this in sigma?
>
Yes.
> or e/A**3?
>
No, but that's an option to request.
> Is this the average density in some sphere surrounding the atom
>
No, see above.
> (like SFALL/OVERLAPMAP)?
>
Hm.. I don't know , I've never used these programs. -:(
Perhaps these are exactly the numbers that I would obtain from
> phenix.map_value_at_point, in which case I've already got what I want.
>
I guess, we can get the numbers you want if we spend a few more iterations
of communicating (may be off-list at this point).
All the best!
Pavel.
Re: [ccp4bb] phenix.real_space_correlation vs overlapmap
On Mon, May 2, 2011 at 2:18 PM, James M Holton wrote: > The CCP4 program that makes the "label map" you are looking for is SFALL. > It can be told to make a .map file where each grid point is still a > floating-point number, but instead of the usual electron density it > "encodes" the residue number, atom number, etc. The OVERLAPMAP program knows > how to decode this, and you specify it as a third map, in addition to the > two you want to correlate. > > However, the distance from the center of each atom that is still "part of > it" is an interesting question. I think the default is 2.5 A or so, but this > depends on the B factor, and probably grid points far from the center of an > atom "should" count less than ones near the middle? > Yes it makes sense that the weighting function for the integration would look like a spherical (or ellipsoidal) gaussian with a variance equal to the sum of the atomic MSD and form factor. > > If what you really want is the number of electrons in a region, then I > would recommend occupancy refinement (which you can now do in REFMAC), and > then adding up occ*Z for the atoms of interest (where Z is the atomic > number). This has the nice property of being independent of F000 estimates. > The tricky part here is that I'd like to do this for density that is poorly-modeled by a single atom. An appropriate model might be a sum over ~10 different atomic positions each with its own B-factor and occupancy, but good luck getting someone to believe you. Since all I'm really interested in is "how much stuff" is there, I don't need to actually convince anyone that my occupancy/B-factor refinement is justified. F000 is going to make this whole thing suspect anyway. What I really want is an END map. Have you got code somewhere? I wasn't planning on sleeping tonight anyway... > > -James Holton > MAD Scientist > > On May 2, 2011, at 11:44 AM, Matt Warkentin wrote: > > > Hi all > > > > I'm trying to 'measure' the density in a region of my structure by > integrating the electron density there (notwithstanding errors in F000). I > understand that both overlapmap and phenix.real_space_correlation compute > density surrounding individual atoms for their calculations. Is there any > way to get that info out of either of them? > > > > Are the per-residue "Fo" and "Fc" columns in the output of > phenix.real_space_correlation actually what I am looking for? If so what > are the units and how is F000 handled? > > > > Thanks a bunch to anyone who can clear this up. > > > > Matt >
Re: [ccp4bb] phenix.real_space_correlation vs overlapmap
On Mon, May 2, 2011 at 12:53 PM, Pavel Afonine wrote: > Hi Matt, > > > I'm trying to 'measure' the density in a region of my structure by >> integrating the electron density there (notwithstanding errors in F000). I >> understand that both overlapmap and phenix.real_space_correlation compute >> density surrounding individual atoms for their calculations. Is there any >> way to get that info out of either of them? >> > > do you want to get electron density maps used in calculation of map CC? If > so, then phenix.maps can compute any map defined as [p][m]Fo+[q][D]Fc, where > p and q are any user-defined numbers. For example: 2Fo-Fc, 2mFo-DFc, > 3Fo-2Fc, Fo-Fc, mFo-DFc, Fc, Fo etc... > What I want is the density 'inside' each atom in the map, so that I can make quantitative comparisons of "how much stuff" is present in a blob (modeled or not). > > > Are the per-residue "Fo" and "Fc" columns in the output of >> phenix.real_space_correlation actually what I am looking for? > > > I'm not sure what you mean... Could you please explain? If you want to get > electron density value (in sigma or e/A**3) computed at each atomic center > (or any specified point in space), then phenix.map_value_at_point would do > it. > Well the output of phenix.real_space_correlation has 'Fc' and 'Fo' columns like so: i_seq : PDB_string element occ b CC Fc Fo No.Points FLAG 0 : " N SER C1002 " N 1.00 28.79 0.9871 2.58 2.75 52 1 : " CA SER C1002 " C 1.00 29.56 0.9945 2.67 2.58 58 2 : " C SER C1002 " C 1.00 31.68 0.9966 3.28 3.22 56 3 : " O SER C1002 " O 1.00 28.79 0.9971 3.51 3.55 55 4 : " CB SER C1002 " C 1.00 36.38 0.9531 2.06 1.80 56 Does the 'Fc' column mean that 2.58 is the average density in atom number 0 (based on F_calc)? If so is this in sigma? or e/A**3? Is this the average density in some sphere surrounding the atom (like SFALL/OVERLAPMAP)? Perhaps these are exactly the numbers that I would obtain from phenix.map_value_at_point, in which case I've already got what I want. > > > >> If so what are the units and how is F000 handled? >> > > It is not included. > > Pavel. > >
Re: [ccp4bb] phenix.real_space_correlation vs overlapmap
The CCP4 program that makes the "label map" you are looking for is SFALL. It can be told to make a .map file where each grid point is still a floating-point number, but instead of the usual electron density it "encodes" the residue number, atom number, etc. The OVERLAPMAP program knows how to decode this, and you specify it as a third map, in addition to the two you want to correlate. However, the distance from the center of each atom that is still "part of it" is an interesting question. I think the default is 2.5 A or so, but this depends on the B factor, and probably grid points far from the center of an atom "should" count less than ones near the middle? If what you really want is the number of electrons in a region, then I would recommend occupancy refinement (which you can now do in REFMAC), and then adding up occ*Z for the atoms of interest (where Z is the atomic number). This has the nice property of being independent of F000 estimates. -James Holton MAD Scientist On May 2, 2011, at 11:44 AM, Matt Warkentin wrote: > Hi all > > I'm trying to 'measure' the density in a region of my structure by > integrating the electron density there (notwithstanding errors in F000). I > understand that both overlapmap and phenix.real_space_correlation compute > density surrounding individual atoms for their calculations. Is there any > way to get that info out of either of them? > > Are the per-residue "Fo" and "Fc" columns in the output of > phenix.real_space_correlation actually what I am looking for? If so what are > the units and how is F000 handled? > > Thanks a bunch to anyone who can clear this up. > > Matt
Re: [ccp4bb] phenix.real_space_correlation vs overlapmap
Hi Matt, I'm trying to 'measure' the density in a region of my structure by > integrating the electron density there (notwithstanding errors in F000). I > understand that both overlapmap and phenix.real_space_correlation compute > density surrounding individual atoms for their calculations. Is there any > way to get that info out of either of them? > do you want to get electron density maps used in calculation of map CC? If so, then phenix.maps can compute any map defined as [p][m]Fo+[q][D]Fc, where p and q are any user-defined numbers. For example: 2Fo-Fc, 2mFo-DFc, 3Fo-2Fc, Fo-Fc, mFo-DFc, Fc, Fo etc... Are the per-residue "Fo" and "Fc" columns in the output of > phenix.real_space_correlation actually what I am looking for? I'm not sure what you mean... Could you please explain? If you want to get electron density value (in sigma or e/A**3) computed at each atomic center (or any specified point in space), then phenix.map_value_at_point would do it. > If so what are the units and how is F000 handled? > It is not included. Pavel.
[ccp4bb] phenix.real_space_correlation vs overlapmap
Hi all I'm trying to 'measure' the density in a region of my structure by integrating the electron density there (notwithstanding errors in F000). I understand that both overlapmap and phenix.real_space_correlation compute density surrounding individual atoms for their calculations. Is there any way to get that info out of either of them? Are the per-residue "Fo" and "Fc" columns in the output of phenix.real_space_correlation actually what I am looking for? If so what are the units and how is F000 handled? Thanks a bunch to anyone who can clear this up. Matt
Re: [ccp4bb] SHELXL/ BAVERAGE
Dear George, Thanks for your effort. That will be of great help! Please let us know about the new version of SHELXL when launched. All the best, Fátima Em Mon, 2 May 2011 17:12:34 +0200 "George M. Sheldrick" escreveu: Dear Fatima, Thanks. I am anyway planning to allow chain identifiers in the next version of SHELXL. Best wishes, George Prof. George M. Sheldrick FRS Dept. Structural Chemistry, University of Goettingen, Tammannstr. 4, D37077 Goettingen, Germany Tel. +49-551-39-3021 or -3068 Fax. +49-551-39-22582 On Mon, 2 May 2011, Fátima Fonseca wrote: Dear George, I really appreciate your kind offer but I have just solved the problem. To all that may have the same issue in the future, I should say that in my case the problem was related to the use of SHELXPRO. I was not using it properly to transform the chains read by SHELXL (numbered) to chains A, B . read by the programs of the CCP4 suite. I was using the G option but I didn't know how to rename the chains. To do that just use the $ symbol and then you'll be asked the chain ID (type A, B, C ), which residues you want to include in that chain and the number of the first residue in the new chain. Do this for all the chains you have in your model. I hope this can help other newbies like me :D Cheers, Fátima Em Mon, 2 May 2011 16:34:50 +0200 "George M. Sheldrick" escreveu: > > Dear Fatima, > > Maybe one of the CCP4 experts has been able to answer >your question. In case this has not happened, I have attached a little >(linux) program >that should do what you want. To find out how to run it, enter the program > name without any parameters: > > ./b-mean > > I have also attached the Fortran source, you will see >that it is a very > simple-minded program. > > Best wishes, George > Prof. George M. Sheldrick FRS > Dept. Structural Chemistry, > University of Goettingen, > Tammannstr. 4, > D37077 Goettingen, Germany > Tel. +49-551-39-3021 or -3068 >Fax. +49-551-39-22582 > > > On Wed, 27 Apr 2011, Fatima Fonseca wrote: > > > Dear all: > > > > I am dealing with a SHELXL/ BAVERAGE issue. > > > > I have refined a few structures in SHELXL but I think I > >am not converting properly the coordinates into the .pdb format. I am > >trying to use Baverage but I get an error > > > > The program run with command: baverage XYZIN > >"C:/datasets/34_MolP.pdb" XYZOUT "C:/datasets/34_MolP_baverage1.pdb" RMSTAB > >"C:/datasets/34_MolP_bavrmstab1.txt" > > has failed with error message > > BAVERAGE: XYZOPEN: Error opening logical name XYZIN > > > > Thank you very much. Any comments will be appreciated. > > > > Fátima > > > >
Re: [ccp4bb] SHELXL/ BAVERAGE
Dear Fatima, Thanks. I am anyway planning to allow chain identifiers in the next version of SHELXL. Best wishes, George Prof. George M. Sheldrick FRS Dept. Structural Chemistry, University of Goettingen, Tammannstr. 4, D37077 Goettingen, Germany Tel. +49-551-39-3021 or -3068 Fax. +49-551-39-22582 On Mon, 2 May 2011, Fátima Fonseca wrote: > Dear George, > > I really appreciate your kind offer but I have just solved the problem. > > To all that may have the same issue in the future, I should say that in my > case the problem was related to the use of SHELXPRO. I was not using it > properly to transform the chains read by SHELXL (numbered) to chains A, B > . read by the programs of the CCP4 suite. > > I was using the G option but I didn't know how to rename the chains. To do > that just use the $ symbol and then you'll be asked the chain ID (type A, B, C > ), which residues you want to include in that chain and the number of the > first residue in the new chain. Do this for all the chains you have in your > model. > > I hope this can help other newbies like me :D > > Cheers, > Fátima > > Em Mon, 2 May 2011 16:34:50 +0200 > "George M. Sheldrick" escreveu: > > > > Dear Fatima, > > > > Maybe one of the CCP4 experts has been able to answer > >your question. In case this has not happened, I have attached a little > >(linux) program > >that should do what you want. To find out how to run it, enter the program > > name without any parameters: > > > > ./b-mean > > > > I have also attached the Fortran source, you will see > >that it is a very > > simple-minded program. > > > > Best wishes, George > > Prof. George M. Sheldrick FRS > > Dept. Structural Chemistry, > > University of Goettingen, > > Tammannstr. 4, > > D37077 Goettingen, Germany > > Tel. +49-551-39-3021 or -3068 > >Fax. +49-551-39-22582 > > > > > > On Wed, 27 Apr 2011, Fatima Fonseca wrote: > > > > > Dear all: > > > > > > I am dealing with a SHELXL/ BAVERAGE issue. > > > > > > I have refined a few structures in SHELXL but I think I > > >am not converting properly the coordinates into the .pdb format. I am > > >trying to use Baverage but I get an error > > > > > > The program run with command: baverage XYZIN > > >"C:/datasets/34_MolP.pdb" XYZOUT "C:/datasets/34_MolP_baverage1.pdb" RMSTAB > > >"C:/datasets/34_MolP_bavrmstab1.txt" > > > has failed with error message > > > BAVERAGE: XYZOPEN: Error opening logical name XYZIN > > > > > > Thank you very much. Any comments will be appreciated. > > > > > > Fátima > > > > > > > > --- > Maria Fátima da Fonseca > PhD student > CESAM & Dep. Biologia > Campus Universitário de Santiago, Universidade de Aveiro > 3810-Aveiro, Portugal > Fax: 234 372 587 > E-mail: ffons...@ua.pt > http://biomicrolab.web.ua.pt > >
Re: [ccp4bb] SHELXL/ BAVERAGE
Dear George, I really appreciate your kind offer but I have just solved the problem. To all that may have the same issue in the future, I should say that in my case the problem was related to the use of SHELXPRO. I was not using it properly to transform the chains read by SHELXL (numbered) to chains A, B . read by the programs of the CCP4 suite. I was using the G option but I didn't know how to rename the chains. To do that just use the $ symbol and then you'll be asked the chain ID (type A, B, C ), which residues you want to include in that chain and the number of the first residue in the new chain. Do this for all the chains you have in your model. I hope this can help other newbies like me :D Cheers, Fátima Em Mon, 2 May 2011 16:34:50 +0200 "George M. Sheldrick" escreveu: Dear Fatima, Maybe one of the CCP4 experts has been able to answer your question. In case this has not happened, I have attached a little (linux) program that should do what you want. To find out how to run it, enter the program name without any parameters: ./b-mean I have also attached the Fortran source, you will see that it is a very simple-minded program. Best wishes, George Prof. George M. Sheldrick FRS Dept. Structural Chemistry, University of Goettingen, Tammannstr. 4, D37077 Goettingen, Germany Tel. +49-551-39-3021 or -3068 Fax. +49-551-39-22582 On Wed, 27 Apr 2011, Fatima Fonseca wrote: Dear all: I am dealing with a SHELXL/ BAVERAGE issue. I have refined a few structures in SHELXL but I think I am not converting properly the coordinates into the .pdb format. I am trying to use Baverage but I get an error The program run with command: baverage XYZIN "C:/datasets/34_MolP.pdb" XYZOUT "C:/datasets/34_MolP_baverage1.pdb" RMSTAB "C:/datasets/34_MolP_bavrmstab1.txt" has failed with error message BAVERAGE: XYZOPEN: Error opening logical name XYZIN Thank you very much. Any comments will be appreciated. Fátima --- Maria Fátima da Fonseca PhD student CESAM & Dep. Biologia Campus Universitário de Santiago, Universidade de Aveiro 3810-Aveiro, Portugal Fax: 234 372 587 E-mail: ffons...@ua.pt http://biomicrolab.web.ua.pt
Re: [ccp4bb] reproducibility of protein crystals
Anita, Proteolysis and oxydation are the most common alteration affecting proteins during the course of crystallization. If you have drops of the trays that yielded crystals I would run a gel on those drops and look at the aspect of protein still around in the drop. That would give you some clues. If there was no reducing agent in the drops I would run a gel with two samples (with and without a reducing agent such as DTT or beta-mercaptoethanol for example). you could also, if you had crystals to spare (although from what you say it does not seem the case), run a gel on a crystal (it takes a little bit of practice) to characterize what is in your crystal or if you have a mass spec at hand look at the content of a crystal. How long did those crystals take to grow? Is there a skin covering your drops? Hope this helps -- Pascal F. Egea, PhD Assistant Professor UCLA, David Geffen School of Medicine Department of Biological Chemistry 314 Biomedical Sciences Research Building office (310)-983-3515 lab (310)-983-3516 email pe...@mednet.ucla.edu
Re: [ccp4bb] reproducibility of protein crystals
On Mon, 2011-05-02 at 15:48 +0800, anita p wrote: > What are the reasons behind nonreproducibility of protein crystals? There are many possible reasons. You could start by checking you protein stock for signs of degradation and preparing it afresh. -- "Hurry up before we all come back to our senses!" Julian, King of Lemurs
[ccp4bb] reproducibility of protein crystals
Dear Crystallographers, I have got my protein crystallized once and then it is not reproducing, though I am using the same batch of protein and same condition. What are the reasons behind nonreproducibility of protein crystals? I am very new to this field hence I apologize if it is a lame question. With regards Anita
