Hi,
Does modeller provide a means of generating models that differ only slightly from an input template?
I have been using modeller to generate models of an antibody, and then docking those models to the solved structure of the target antigen protein for that antibody. Multiple models are required, as the docking is very sensitive to the correct conformation of loops at the docking interface. I have some resulting docked conformations that are receiving pretty high ZRANK scores, but upon inspection, it seems that they could all be just a little bit better fit. I was hoping I could feed these docked conformations as templates into modeller, and generate a large number of very similar models, with slightly differing energies.
I'm already comfortable generating input PIR alignments with multiple chains; I'm just interested in hearing what the recommended approach would be for such a refinement process.
Cordially, Jake
Jake Gunn-Glanville wrote: > Does modeller provide a means of generating models that differ only > slightly from an input template?
Sure - you could build some models using a simple 1:1 alignment to the template. The final models will look very similar to the input (by construction) but you should expect to see some variability, particularly in the regions that are not strongly restrained (e.g. sidechains). You could add in some extra restraints around the docked region, if you have some extra data or intuition about the interactions. But Modeller is designed to generate good solutions to your scoring function, not to sample from an ensemble - if you want to do that you might have more luck with an MD package.
Ben Webb, Modeller Caretaker