After a nightmare of MR trials, I've finally got a resonable match using poly-ala model by first pruning (gedit) then chainsaw. Any recommendations of a good site for the next steps would be helpful to me. I know I have to mutate all residues back as well as build the remaining missing residues from the pruning. Your thoughts on what I should be looking for and how I know if the results are good and worth refining would be helpful also.
The next step is to replace the residues to the correct side chains. It is very easy to get register errors (moving sections of your structure one residue forward or backward). First of all you can replace big, bulky side chains in well resolved regions - the interior of the protein, between helicies or sheets, rather than at the flexible loops. Tryptophan, tyrosine and phenylalanine are best for this, but if you have arginine, isoleucine or glutamate in interior regions these can also be helpful, as the densities are easily recognisable.
One you have a few residues you are absolutly confident of near each other you can mutate between these and get a good section of your protein. If you can now be sure of around 50% of your structure is in the right places you can then use this as a template to automatically build again and this should give you some better links between residues. Delete any loops that are poor and rebuild these, as alanines, re-refine and see if you can now see side chains. Remeber some side chains will noto resolve on the surface your protein so remain as alanine during refinement (you can set atom ocupancy to 0 so it is the correct residue, but not included in the model). If the R values are getting better then you are improving your structures, and MolProbity will help to tell you hw reliable your structure is.
I hope that helps,
Ellis
The next step is to replace the residues to the correct side chains. It is very easy to get register errors (moving sections of your structure one residue forward or backward). First of all you can replace big, bulky side chains in well resolved regions - the interior of the protein, between helicies or sheets, rather than at the flexible loops. Tryptophan, tyrosine and phenylalanine are best for this, but if you have arginine, isoleucine or glutamate in interior regions these can also be helpful, as the densities are easily recognisable.
One you have a few residues you are absolutly confident of near each other you can mutate between these and get a good section of your protein. If you can now be sure of around 50% of your structure is in the right places you can then use this as a template to automatically build again and this should give you some better links between residues. Delete any loops that are poor and rebuild these, as alanines, re-refine and see if you can now see side chains. Remeber some side chains will noto resolve on the surface your protein so remain as alanine during refinement (you can set atom ocupancy to 0 so it is the correct residue, but not included in the model). If the R values are getting better then you are improving your structures, and MolProbity will help to tell you hw reliable your structure is.
I hope that helps,
Ellis
Ellis got it right on, but I'll add some comments on validation and what to look for based on what you described.
If MR has been difficult, then you want further confirm that your solution is correct. I'm going to guess your free R is currently 0.3 to 0.4, and your electron density follows the chain but is broken and messy. At this point your phases are poor, which means initial interoperation will be tough, but you will see huge improvements in free R as you refine and regenerate maps. Try doing one round of restrained refinement on the polyA model (very restrained: use secondary structure restraints in phenix.refine or jelly body restraints in refmac5, omit B-factor refinements). Look for a big drop in R free. Look for improved density connectivity.
Auto building and phase improvement software have gotten remarkably good, even at medium to low resolution. Try the automatic model building modules in PHENIX or Buccaneer in CCP4 and see what you get. You might find that they could rebuild most or all of the structure for you. At the very least it will help with the initial sequence registry that Ellis outlined.
If initial refinement and model building isn't causing noticeable improvements in statistics or maps, I'd be wary of your molecular replacement solution. I'd reconsider your space group if it has apparent high symmetry. Sometimes different subunits of oligomers are in slightly different conformations, and in low quality datasets might appear as crystallographic symmetry.
Actually, the first and simplest thing you might try is morphing the molecular replacement solution to best fit the data. There's a nice implementation of that in Phenix and the manual is here: http://www.phenix-online.org/documentation/reference/morph_model.html
What Justin and Ellis said is true but I just wanted to add some links to get you started:
Buccaneer, model autobuilding: http://ccp4wiki.org/~ccp4wiki/wiki/index.php?title=Automated_model_building_with_Buccaneer
ARP/wARP, model autobuilding: http://www.embl-hamburg.de/ARP/
phenix.autobuild, model autobuilding: http://www.phenix-online.org/documentation/reference/autobuild_gui.html
You could also use AMPLE or phenix.rosetta to combine your phase information with homology modelling or SHELXE to do density modification with autotracing to extend your poor phases.
AMPLE: http://ccp4wiki.org/~ccp4wiki/wiki/index.php?title=Ab_initio_modelling_and_automated_molecular_replacement_with_AMPLE
phenix.mr_rosetta: http://www.phenix-online.org/documentation/reference/mr_rosetta.html
SHELXE: http://www.ncbi.nlm.nih.gov/pubmed/24189237
Morten
- Badges
- Science topic
- Similar topics
- Chemistry
- Biochemistry