I'm currently refining an enzyme structure (data from X-ray crystallography). Resolution is 2.8 Å; model was generated using MR on a template with 62% sequence identity and Autobuild (Phenix). I'm able to get good R-factor (0.218) and R-free (0.256) values, but the Ramachandran plot shows 95.5% favoured, 3% allowed and 1.5% generously allowed. There are a few rotamer outliers too. Is this acceptable for 2.8 Å resolution structure? I'm not able to manually fix the outliers using Coot. Are there any strategies to improve the geometry?
1.5% generously allowed is OK if they mainly consists of Prolines and Glycines. Main objective of refining is to fit the model perfect in harmony with the obtained electron density map. I think you should be more concerned about quality of your model in terms of geometry than worrying more about Rfactors and Rfree. You may get better R values in case of model bias but that won't solve the right purpose. Yes R values tells you the quality of structure but this is not the only criteria for a good structure. In many cases like twinning or NCS you may get higher R value in spite of having high resolution structure.
You may also try TLS or NCS to see the better result.
You should also consider Rms Bound and Angle values. Those should be lower than let's say 0.011A and 1.5° respectivelly.
Are you doing "real space refinement" in Phenix ?
What do you mean by not being able to manually fix the outliers in coot ?
RMS bond and angle are average values so if he has a few local areas with incorrect modelling then that might not show up.
Is it in a loop with poor density that you get these Ramachandran "outliers"? Are the tricky ones clumped together or evenly spread? Did you use Ramachandran restraints during your refinement (a big no-no)? What happens if you throw it into ensemble refine (http://phenix-online.org/documentation/reference/ensemble_refinement.html)? What happens if you delete the problem areas and about five residues before and after and try to autobuild the structure again? Add TLS to your model and your R-factors should go down but I don't know how much it will change your maps.
Sometimes a protein is simply in a strained conformation and rotamer outliers can be indicative of that. It is always easier to help someone when you are actually sitting with the structure and the electron density maps so if you can find someone in your department who knows what they are doing and can spare you a couple of hours then I would highly recommend that.
Thanks everyone for the replies. @Morten: Most of the 'outliers' are in a 15 amino acid loop that had very poor density. My protein is a tetramer, and there are 3-4 tricky residues evenly spread in each subunit. I have not used Ramachandran restraints, only NCS restraints during the refinement. I'll try deleting the problem areas and autobuild again.
Okay, I don't think you are going to get a satisfactory answer then. There will be multiple correct solutions but the density will be from where some of those solutions overlap. When the refinement software then tries to fit the electron density and the single model together then you get less than ideal geometry.
You should get a collection of good solutions with the ensemble refinement. But you can't deposit that. So your options are to delete the residues with poor density and submit without them or accept that your geometry is poor there. Either way you will have to explain in your paper what you chose and why. I think deleting them was less frowned upon in the old days. Even though you should always delete N- and C-terminal residues that you don't have density for.
I've reduced the number of Ramachandran outliers to 1.2% (15 residues, a few prolines and others from the loop with bad density). Favoured = 95.4%, Allowed = 3.4%. Will this be enough?
seems ok, 2.8 A isn't bad. some proteins are submitted at 3.3 and 3.4 A, and with R factor near 0.3! really depends what you are after, and what is in the structure missing or unclear, what region exactly, is there surface loop not good, etc. If RCSB PDB tools accept it just do it if you can't get better, also following comments above.
- Badges
- Science method
- Similar topics
- Methodology
- Crystallography
- X-ray Diffraction