Hi!
I am looking for a simple method to score a model from Modeller directly against raw experimental cristallography data. The idea is to use models from Modeller for molecular replacement in borderline situations - that is in situations where the standard cristallographers molecular replacement tools fail. Thus, the method should be automatic and not require hand-tuning, so that I can generate lots of different models (e.g. based on different loop refinement parameters) until I find one that does the trick. Do you know about someone who already tried this.
Thank you very much,
kind regards, Karsten.
Hello,
In case of loop modeling you need to identify the region(s) where you want to apply it, but after that -because it is an ab initio approach- all the generated models are individually optimized samples. Probably that is what you are looking for. It is automatic, beside identifying the segment, you do not need to do anything else.
Andras
Karsten Suhre wrote: > > Hi! > > I am looking for a simple method to score a model from Modeller directly > against raw experimental cristallography data. The idea is to use models from > Modeller for molecular replacement in borderline situations - that is in > situations where the standard cristallographers molecular replacement tools > fail. Thus, the method should be automatic and not require hand-tuning, so > that I can generate lots of different models (e.g. based on different loop > refinement parameters) until I find one that does the trick. Do you know > about someone who already tried this. > > Thank you very much, > > kind regards, Karsten. > > -- > Karsten Suhre > Information Genetique & Structurale > UMR CNRS 1889 > 31, chemin Joseph Aiguier > F-13402 Marseille Cedex 20 > mailto:Karsten.Suhre@igs.cnrs-mrs.fr > http://igs-server.cnrs-mrs.fr/~suhre
Right - and what I am seeking is a way of telling Modeller which loops to refine on the basis of christallographic data, or maybe another way to constrain Modeller with christallographic data. I suppose there are a lot of possible ways to do this by hand, depending on the tools you have, but as I am still quite new to this domain, I was hoping to find someone who has some experience with this in order not to reinvent the wheel.
Kind regards, Karsten.
On Tuesday 27 August 2002 18:04, you wrote: > Hello, > > In case of loop modeling you need to identify the region(s) where you > want to apply it, but after that -because it is an ab initio approach- > all the generated models are individually optimized samples. Probably > that is what you are looking for. > It is automatic, beside identifying the segment, you do not need to do > anything else. > > > > Andras > > Karsten Suhre wrote: > > Hi! > > > > I am looking for a simple method to score a model from Modeller directly > > against raw experimental cristallography data. The idea is to use models > > from Modeller for molecular replacement in borderline situations - that > > is in situations where the standard cristallographers molecular > > replacement tools fail. Thus, the method should be automatic and not > > require hand-tuning, so that I can generate lots of different models > > (e.g. based on different loop refinement parameters) until I find one > > that does the trick. Do you know about someone who already tried this. > > > > Thank you very much, > > > > kind regards, Karsten. > > > > -- > > Karsten Suhre > > Information Genetique & Structurale > > UMR CNRS 1889 > > 31, chemin Joseph Aiguier > > F-13402 Marseille Cedex 20 > > mailto:Karsten.Suhre@igs.cnrs-mrs.fr > > http://igs-server.cnrs-mrs.fr/~suhre
These are both possible. I would write a script that automatically generates modeller top files. The top file will have a definition to pick and model loops that are not well refined in the crystallographic map. You would need a definition how to identify these regions from your experimental data.->this is a more ab initio approach
Alternatively, you could define a set of "user defined retsraints" from crystallographic data in general (you will find more about it in the manual in teh FAQ section). In this second case I would just generate many models without loop modeling, maybe putting the starting conformations in a sufficiently "loose" state so you can better explore and optimize your models ("SET DEVIATION = 6" i.e. in angstroms). ->this is a more comparative modeling approach
Both approaches assume you have some kind of rough initial model (the well defined parts of your structure)
However I have not done it, if that is what you are asking for.
best, Andras
Karsten Suhre wrote: > > Right - and what I am seeking is a way of telling Modeller which loops to > refine on the basis of christallographic data, or maybe another way to > constrain Modeller with christallographic data. I suppose there are a lot of > possible ways to do this by hand, depending on the tools you have, but as I > am still quite new to this domain, I was hoping to find someone who has some > experience with this in order not to reinvent the wheel. > > Kind regards, Karsten. > > On Tuesday 27 August 2002 18:04, you wrote: > > Hello, > > > > In case of loop modeling you need to identify the region(s) where you > > want to apply it, but after that -because it is an ab initio approach- > > all the generated models are individually optimized samples. Probably > > that is what you are looking for. > > It is automatic, beside identifying the segment, you do not need to do > > anything else. > > > > > > > > Andras > > > > Karsten Suhre wrote: > > > Hi! > > > > > > I am looking for a simple method to score a model from Modeller directly > > > against raw experimental cristallography data. The idea is to use models > > > from Modeller for molecular replacement in borderline situations - that > > > is in situations where the standard cristallographers molecular > > > replacement tools fail. Thus, the method should be automatic and not > > > require hand-tuning, so that I can generate lots of different models > > > (e.g. based on different loop refinement parameters) until I find one > > > that does the trick. Do you know about someone who already tried this. > > > > > > Thank you very much, > > > > > > kind regards, Karsten. > > > > > > -- > > > Karsten Suhre > > > Information Genetique & Structurale > > > UMR CNRS 1889 > > > 31, chemin Joseph Aiguier > > > F-13402 Marseille Cedex 20 > > > mailto:Karsten.Suhre@igs.cnrs-mrs.fr > > > http://igs-server.cnrs-mrs.fr/~suhre
These are both possible. I would write a script that automatically generates modeller top files. The top file will have a definition to pick and model loops that are not well refined in the crystallographic map. You would need a definition how to identify these regions from your experimental data.->this is a more ab initio approach
Alternatively, you could define a set of "user defined retsraints" from crystallographic data in general (you will find more about it in the manual in teh FAQ section). In this second case I would just generate many models without loop modeling, maybe putting the starting conformations in a sufficiently "loose" state so you can better explore and optimize your models ("SET DEVIATION = 6" i.e. in angstroms). ->this is a more comparative modeling approach
Both approaches assume you have some kind of rough initial model (the well defined parts of your structure)
However I have not done it, if that is what you are asking for.
best, Andras
Karsten Suhre wrote: > > Right - and what I am seeking is a way of telling Modeller which loops to > refine on the basis of christallographic data, or maybe another way to > constrain Modeller with christallographic data. I suppose there are a lot of > possible ways to do this by hand, depending on the tools you have, but as I > am still quite new to this domain, I was hoping to find someone who has some > experience with this in order not to reinvent the wheel. > > Kind regards, Karsten. > > On Tuesday 27 August 2002 18:04, you wrote: > > Hello, > > > > In case of loop modeling you need to identify the region(s) where you > > want to apply it, but after that -because it is an ab initio approach- > > all the generated models are individually optimized samples. Probably > > that is what you are looking for. > > It is automatic, beside identifying the segment, you do not need to do > > anything else. > > > > > > > > Andras > > > > Karsten Suhre wrote: > > > Hi! > > > > > > I am looking for a simple method to score a model from Modeller directly > > > against raw experimental cristallography data. The idea is to use models > > > from Modeller for molecular replacement in borderline situations - that > > > is in situations where the standard cristallographers molecular > > > replacement tools fail. Thus, the method should be automatic and not > > > require hand-tuning, so that I can generate lots of different models > > > (e.g. based on different loop refinement parameters) until I find one > > > that does the trick. Do you know about someone who already tried this. > > > > > > Thank you very much, > > > > > > kind regards, Karsten. > > > > > > -- > > > Karsten Suhre > > > Information Genetique & Structurale > > > UMR CNRS 1889 > > > 31, chemin Joseph Aiguier > > > F-13402 Marseille Cedex 20 > > > mailto:Karsten.Suhre@igs.cnrs-mrs.fr > > > http://igs-server.cnrs-mrs.fr/~suhre
-
Andras Fiser -
Karsten Suhre