Date: 19 November 2011 12:09
Dear All,
We have an anisotropic dataset of 3.3 A and it was solved (not by me) with P6522 with R/freeR 29.1/37.3.
I got the corrected mtz file by plugging in the .HKL (P6122) file to anisotropy diffraction server at 2.04 A. I reindexed this p6122 to p6522 and extended the resolution and refined (refmac) the structure to R/freeR 36.40/38.50. With aotoncs option, fixing all Ramachnadran and rotamer outliers I got it 30/32. When I added waters and it went down to 27.5/31.2. At this point I recognized that my new .mtz file from anisotropy server has different R flag than the earlier one (3.3A) so I copied the R flag and did refinemnt to get R/Rfree 0.2682/0.3247. When I looked at the refined structure I found more outliers than I fixed in earlier round. I did fix all the outliers and without NCS and waters it gives R/Rfree 0.2906/0.3325. At all the stages I look at outliers at molprobity server which suggested structure is 10th percentile and after refinement more outliers comes back. At stage-1 map looked far better so was happy that anisotropy correction has worked for me (this was my first time handling this type of dataset) but further refinement didn't make it look any better.I use both refmac and autoBuster for refinement. http://www.flickr.com/photos/rajesh_ccp4/sets/72157628048657095/
This protein is an human enzyme and a bacterial homologue which has 38% identity has been used to solve the Apo structure (2.7 A, pC2221, R/freeR 23.03/27.96, molprobity is around 50th percentile). I looked in to this I try to fix all the outliers and try to improve molprobity score but it just refused to improve as after refinement I get more outliers. This Apo structure was used to solve the mutant structure at 3.3 A. I believe that both structure could have better R/freeR and excellent molprobity scores than what they have now. I am not able to recognise if there is any problem in Apo structure and if errors have come to mutant so both of them refuse to improve.
I wondered if there is any model bias (I don't know if it's the case but nothing was coming to my mind) so thought using ARP/wARP classic to build model from existing model but it complained that "The wilson plot is very bad and ARp/wARP is very unlikely to run in a sensible way. Please check your data" . http://www.flickr.com/photos/rajesh_ccp4/sets/72157628048687955/
At this point I dont know how to systematically dissect this problem. I know there could be wrong in several places but with my only '2-3 structure experience' I am not able to identify the regions to look for error but I think something is not right. I really appreciate if you give me some suggestions/ideas/directions/tips so that I could recognize problem and improve structure and learn some more.
I appreciate your valuable time.
Regards,
Rajesh
----------
From: Eleanor Dodson
I think you are proving yet again that refinement at 3.3A is not easy.
Indeed there are probably multiple conformations for parts of the structure and that may well be why your data is at low resolution and anisotropic. Maybe this is the best you can do..
I think I would make sure the apo structure is as good as it can be, then fit that to the 3.3A data set, and only use that 3.3A data to deduce whatever features differ from the APO structure.
Eleanor
----------
From: Rajesh kumar
----------
From: Eleanor Dodson
It is hard to make sensible comments without seeing your maps.
In these cases I use coot to validate.
Run the ramachandran check - click on all outliers and see if what the electron density looks like. Sometimes you do see indicators on how to make corrections. Sometimes you have a CIS peptide which has been forced to be TRANS - that can cause disturbances in neighbouring residues too.
For the clashes - I use the REFMAC listing to see where they are - again check the density - and rebuild if possible. Some residues may simply be invisible, then I put the occupancy to 0.00 .
Bad rotamers and angles can be caused by crystal packing but again check..
And remember that in most structures there are poorly ordered regions - if the offending residues are there you probably cant refine them..
2.74A is not high resolution, so you will have problems correcting everything
Eleanor
Clashscore, all atoms: 33.73 50th percentile* (N=185, 2.740Å ± 0.25Å)Geometry Poor rotamers 8.11% Goal:<1%Ramachandran outliers 0.52% Goal:<0.2%Ramachandran favored 95.19% Goal:>98%Cβ deviations>0.25Å 2 Goal: 0MolProbity score^ 3.04 53rd percentile* (N=5278, 2.740Å ± 0.25Å)Residues with bad bonds: 0.00% Goal: 0%Residues with bad angles: 0.34% Goal:<0.1%
If i fix the all the possible outliers and refine the structure it wouldn't improve the molprobity scores. I am not happy about the above scores at all. I am wondering if this is an indication of any wrong in the structure or is this common for an enzyme. I have reprocessed the data to make sure space group is same C2221.Any help would help me understand this and learn more.
No comments:
Post a Comment