Advice on heteroatoms, metals and water
Hi. This is directed to :
>Karsten Suhre Karsten.Suhre@igs.cnrs-mrs.fr (IGS) >Introduction of a Ca binding site
>Moshe Amitay moshe@noamy.ch.biu.ac.il ("Bar-Ilan University, Ramat-Gan, >building a model with water molecules.
From: Tjaart de Beer tjaart@tuks.co.za >Modelling with metals and hetero.
The practical advice : Modeller treats the 'hetatm' entries as residues for much of the modelling process. By this, I mean that it is as picky as with the other residue entries when you describe them in the alignment. You can see all the types of residue entry that it knows about in the mod6v2/lib/restyp.lib file.
Its always simpler to 'SET HETATM_IO', rather than change the pdb file. If modeller knows about the heteroatom residue code then it will read it, and not attempt to attach proximal metal ions to the polypeptide backbone.
I was mislead by the chain break character '/', which actually introduces a chain break for your model, rather than describing a corresponding template (? not really sure here). I got away with this example : 1tem:Template:1: :3: :.:.:. WLMzz*
1hom:Homolog:1: :3: :.:.:. WITzz*
Where I just used set HETATOM_IO on. The important thing is the numbering used in the PIR header for each protein sequence - this only relates to the amino acids, not the hereroatoms.
I imagine that you need to use set WATER_IO on, and put in the waters, too, but the chain break didn't seem to help.
Above all, look carefully at the numbers that Modeller gives you when it complains that you don't have enough residues in the PDB file or the sequence. For Moshe's case, the chain break may well have been the problem, but it can't hurt to use the unedited PDB file but setting HETATM_IO as well as WATER_IO (but maybe you could delete the un-necessary waters and heteroatoms from the pdb file just to make the alignment simpler!).
As for the optimization behaviour with metal atoms - I didn't really notice modeller doing any optimization of the bonding geometry, but in my case these contacts were thoroughly conserved. However, I can't see any parameters in the MD forcefield definitions to indicate that charge/ligand geometry is taken into account, it is probably all done through restraints.
Think thats it - sorry for not giving a concrete example!
j--
ps. For Calcium - the HETATM's residue entry should be CAL or CA (that is, not CA2), and '3' is the alignment symbol used in the PIR file (see the restyp.lib file). _______________________________________________________________________ Dr JB Procter:Biomolecular Modelling at ZBH - Center for Bioinformatics Hamburg http://www.zbh.uni-hamburg.de/mitarbeiter/procter
Hi!
Thank you for this information. In fact, I was hoping that Modeller had some forcefields hidden somewhere in the depth of its code to do loop modelling around Ca2+ binding sites. The real problem is not to place the Ca2+, but to have all the oxygens of the different side chains point its way. My templates are not close enough to get this right, and I have even one case where the Ca-binding site is only present in the target.
In parallel, I try to get the sidechains in place using Amber, but modelling a Ca2+ binding site seems not to be a standard procedure (although this should occur quite often). Another problem with Amber is that I would have to keep the rest of my protein restrained (otherwise my structure would drift away from its crystall-structure status by completely resolving it in water - but I want to use the model for protein crystal structure building later). So the choice of these restrains turns out to be quite tricky as well.
So I am still wondering whether it would not be possible to teach MODELLER how to model a Ca2+ in its loop refinement algorithm. If someone could give me some technical advice on how to do this (or where to find more specific documentation) I would be very grateful.
Kind regards, Karsten.
On Tuesday 19 August 2003 21:04, Jim Procter wrote: > Hi. This is directed to : > >Karsten Suhre Karsten.Suhre@igs.cnrs-mrs.fr (IGS) > >Introduction of a Ca binding site > > > >Moshe Amitay moshe@noamy.ch.biu.ac.il ("Bar-Ilan University, Ramat-Gan, > >building a model with water molecules. > > From: Tjaart de Beer tjaart@tuks.co.za > > >Modelling with metals and hetero. > > The practical advice : > Modeller treats the 'hetatm' entries as residues for much of the modelling > process. By this, I mean that it is as picky as with the other residue > entries when you describe them in the alignment. You can see all the types > of residue entry that it knows about in the mod6v2/lib/restyp.lib file. > > Its always simpler to 'SET HETATM_IO', rather than change the pdb file. If > modeller knows about the heteroatom residue code then it will read it, and > not attempt to attach proximal metal ions to the polypeptide backbone. > > I was mislead by the chain break character '/', which actually introduces a > chain break for your model, rather than describing a corresponding template > (? not really sure here). I got away with this example : > 1tem:Template:1: :3: :.:.:. > WLMzz* > > 1hom:Homolog:1: :3: :.:.:. > WITzz* > > Where I just used set HETATOM_IO on. The important thing is the numbering > used in the PIR header for each protein sequence - this only relates to the > amino acids, not the hereroatoms. > > I imagine that you need to use set WATER_IO on, and put in the waters, too, > but the chain break didn't seem to help. > > Above all, look carefully at the numbers that Modeller gives you when it > complains that you don't have enough residues in the PDB file or the > sequence. For Moshe's case, the chain break may well have been the problem, > but it can't hurt to use the unedited PDB file but setting HETATM_IO as > well as WATER_IO (but maybe you could delete the un-necessary waters and > heteroatoms from the pdb file just to make the alignment simpler!). > > As for the optimization behaviour with metal atoms - I didn't really notice > modeller doing any optimization of the bonding geometry, but in my case > these contacts were thoroughly conserved. However, I can't see any > parameters in the MD forcefield definitions to indicate that charge/ligand > geometry is taken into account, it is probably all done through restraints. > > Think thats it - sorry for not giving a concrete example! > > j-- > > ps. For Calcium - the HETATM's residue entry should be CAL or CA (that is, > not CA2), and '3' is the alignment symbol used in the PIR file (see the > restyp.lib file). > _______________________________________________________________________ > Dr JB Procter:Biomolecular Modelling at ZBH - Center for Bioinformatics > Hamburg http://www.zbh.uni-hamburg.de/mitarbeiter/procter
Karsten wrote : > Thank you for this information. In fact, I was hoping that Modeller had > some forcefields hidden somewhere in the depth of its code to do loop > modelling around Ca2+ binding sites. The real problem is not to place the
no such luck - it uses a reduced version of amber without the bells and whistles (FFT/QMM electrostatics, etc).
> Ca2+, but to have all the oxygens of the different side chains point its > way. My templates are not close enough to get this right, and I have even > one case where the Ca-binding site is only present in the target.
Well. Without good homologs for the metal binding site, which means that the local environment of the loop is roughly similar too, you will have a hard time applying any kind of homology modelling procedure reliably. Particularly so, if you want to use this for molecular replacement. This is because MODELLER uses a knowledge-based algorithm to generate the constraints, and is not necesarily going to be able to predict something it hasn't seen before (ie in the loop-database used in the loop-modelling procedure).
The only 'reliable' approach is to find an ab-initio geometry for the loop, which you might get using a good dynamics package. Amber is perhaps appropriate (IIRC) if you use the proper electrostatics, but there are other packages out there. Either way, you will need a cautious combination of simulation and homology modelling in order to get anywhere. The complex issue of setting restraints on the overall fold of the molecule is unavoidable, I'm afraid.
However, if you just want to make a nice picture using the text-book coordination geometry, you could try adding some appropriate distance and dihedral restraints to your model - have a look here : http://salilab.org/modeller/manual/node113.html
have fun ;-) j.
participants (2)
-
Jim Procter -
Karsten Suhre