Hi all, I have a very basic question: In case of homology modeling, is it preferable to model one domain at a time or the entire structure at once? In other words, when I use a template prediction tool, should I give a query, say 3 times for a 3-domain protein or one query for the entire protein? If the answer is domain wise, then how should the assembly of the entire protein be made, especially in case where multiple templates are used
I guess it all depends...Do you have a homologous protein for which the structure off all three domains has been solved? If you do, you can probably use that to model your protein. If you do not have such a structure, I would do homology modeling on three separate domains. How to reassemble them into one protein is another question. Without good structural information from an homologous protein, I would not try to reassemble these three domains into a single protein. I would try to use the three separate domains to learn as much as I could and hope that someone will solve the real structure soon. I think Rosetta has made some claims about trying to reassemble these domains, but I do not know how well that approach works...
Hi, thanks for replying. The protein that I'm interested in is a parasite protein and I want to concentrate on how similar or different it is from its host. The homologous protein in the host has one domain whose structure is crystallized. However, the template I got was of yeast, which is the only organism with its full length structure determined. I would like to use the host protein as the template if that wouldn't be a bad move
So are there actually three different proteins in question: parasite, host, and yeast? If all three are homologous, but the only full length protein with a structure is for yeast, and the host protein has one domain solved, I would actually construct a homology model for both the host and parasite sequences on the yeast (as long as all three proteins are homologs and not just that they possess homologous domains).
I did as Dan and Whetting have said--modeled the host domains individually on the yeast full length template --and it's making sense to me, so thanks to both! The sequence similarity of all domains is just about 26% for all but structurally, they seem to be okay to start with. However, I'm having problems with superimposing all three domains at once on the template. Although this is possible with the tool I'm using--Discovery Studio 3.5 (standalone) I'm not able to save the superposed file in the .pdb format. Is there an online server to do this (or better still, fix this problem on DS itself)? Pymol installation is right now on hold because of a local server interference, so I'd like an online alternative to do my job until Pymol is up and running again.
Guess I'll keep coming back here until my model is all ready!
This should be posted either as a comment or an update to your original post (and select Whetting's answer above as the accepted answer, instead of a new answer. In response though, I found some success using FATCAT structural alignment (online tool) to superimpose my structures. I use VMD for visualizing instead of PyMol. Although I did find that FATCAT superimposed structures tended to look funny in cartoon format. If you did your homology modeling in Swiss-Prot the models should be in the same coordinate system. You can then just load each PDB file separately in VMD and it will show them all superimposed on top of each other by default. You can then manipulate each representation (an individual PDB file) individually as you wish. VMD has a bit of a steep learning curve but I like it for its power.
Hi, thanks for replying. The protein that I'm interested in is a parasite protein and I want to concentrate on how similar or different it is from its host. The homologous protein in the host has one domain whose structure is crystallized. However, the template I got was of yeast, which is the only organism with its full length structure determined. I would like to use the host protein as the template if that wouldn't be a bad move
So are there actually three different proteins in question: parasite, host, and yeast? If all three are homologous, but the only full length protein with a structure is for yeast, and the host protein has one domain solved, I would actually construct a homology model for both the host and parasite sequences on the yeast (as long as all three proteins are homologs and not just that they possess homologous domains).
I agree, if there are 2 known structures, I would calculate the model based on both. Assuming they are real homologous...
Just to clarify, you want to use the host protein to model the parasite protein? If they are true homologoues that should be fine!