RSR of tetramer, rotamers of different protomers/chains change
Hi All, I want to apologize in advance for such a basic question. I'm working with a tetrameric protein and am attempting to build a model. The resolution estimate of the map determined by cryoEM is 3.8A. Some parts of the map are great while some other parts not so much. To assist with modelling the tetramer, I used the local refinement of one protomer to build a protomer model (one chain). I then used that protomer model as a template to build the full tetrameric model, and I make sure that each protomer is exactly the same (using ChimeraX or coot) by copying one chain over to the next and making certain they are all exact copies of one another. When doing real space refinement in phenix 2.0-5936 of the tetramer model, I end up with good statistics and what I think is a complete model. RCSB/comprehensive cryoEM validation also seem to complement my modelling. However upon closer inspection I noticed that many rotamers across the model are different, for example Chains A and B will have a Gln in one position, while chains C and D will have rotamers in totally different positions. These rotameric seem to happen in substandard regions of map quality. My understanding is that all the protomers in a C4 channel should look exactly the same. To attempt to rectify this problem I've made sure that NQH_flips is off (rotameric changes still happen with many different residues), as well as tried using reference restraints of my protomer, or NCS constraints of one protomer (these options results in a degradation of model quality). I'm missing or misunderstanding something here (or maybe a combo of both), and I was wondering if there was any advice from the community on how to proceed. Kind Regards, Seb
Hi Seb, assuming the map was symmetrized (which is likely at 3.8A), steps are: 0. Get initial model (AphaFold, homologous Xray, etc) 1. Find symmetry of the map; 2. Extract box containing symmetry unique part of the map; 3. Dock your model into the map; 4. Rebuild/refine model in the boxed unique part; 5. Once done building/rebuilding, apply symmetry to the built model to generate the whole tetramer; 6. Clashes are likely at the interface between symmetry copies; 7. Resolve clashes by iterations of manual rebuilding and refinement. Make sure you run refinement with NCS constraints, for the tetramer you will have one master and three dependent NCS copies. This will ensure that all copies are atom content identical and if superposed, their superposition rmsd is exactly zero. If that's not the case -- something must have gone wrong. 8. You are all done -- run final Comprehensive Validation in Phenix GUI and publish/deposit your model. If the map was not symmetrized -- genuine structural differences are expected to be observed between symmetry copies -- do essentially the same steps except at step #7 do NOT use NCS constraints at all during refinement. All 0-9 steps can be done in Phenix GUI or in the command line. This is one of the tutorial items that we demonstrate at most Phenix user workshops. Good luck! Pavel On 7/23/26 14:39, skaruppan--- via phenixbb wrote:
Hi All,
I want to apologize in advance for such a basic question.
I'm working with a tetrameric protein and am attempting to build a model. The resolution estimate of the map determined by cryoEM is 3.8A. Some parts of the map are great while some other parts not so much. To assist with modelling the tetramer, I used the local refinement of one protomer to build a protomer model (one chain). I then used that protomer model as a template to build the full tetrameric model, and I make sure that each protomer is exactly the same (using ChimeraX or coot) by copying one chain over to the next and making certain they are all exact copies of one another. When doing real space refinement in phenix 2.0-5936 of the tetramer model, I end up with good statistics and what I think is a complete model. RCSB/comprehensive cryoEM validation also seem to complement my modelling. However upon closer inspection I noticed that many rotamers across the model are different, for example Chains A and B will have a Gln in one position, while chains C and D will have rotamer s in totally different positions. These rotameric seem to happen in substandard regions of map quality.
My understanding is that all the protomers in a C4 channel should look exactly the same. To attempt to rectify this problem I've made sure that NQH_flips is off (rotameric changes still happen with many different residues), as well as tried using reference restraints of my protomer, or NCS constraints of one protomer (these options results in a degradation of model quality).
I'm missing or misunderstanding something here (or maybe a combo of both), and I was wondering if there was any advice from the community on how to proceed.
Kind Regards, Seb _______________________________________________ phenixbb mailing list -- [email protected] To unsubscribe send an email to [email protected] Unsubscribe: phenixbb-leave@%(host_name)s
Pavel, Thank you very much for your reply, I do appreciate it greatly. I've done steps 0-6. Using the Superpose Models in Phenix I've verified that for my input model, the chains have RMSD values of 0.00 with each other (i.e. A-B is 0, A-C is 0, etc). Then with RSR, I've ensured that the NCS constraints on, and I've made sure that chain 'A' is the master and 'B-D' are the dependents in the selection panel, and I hit run. Unfortunately I'm missing something or there's an issue because I still see rotamer changes between chains. Using the Superpose Models tool in phenix I see that the RMSD between chains is slight, ~0.03 to 0.02. What do you suppose I could be missing? Kind Regards, Seb
Hi Seb, At this point, I need to take a close look at your files to diagnose the issue. Could you please share (off-list) the model before refinement where the RMSD between copies is exactly zero, and the model after refinement where these chains do not overlay perfectly? Also, please send the map, .log, and .eff files. With these files, I’ll try to replicate your results and explain what’s going on. Thanks! Pavel On 7/23/26 16:43, skaruppan--- via phenixbb wrote:
Pavel, Thank you very much for your reply, I do appreciate it greatly.
I've done steps 0-6. Using the Superpose Models in Phenix I've verified that for my input model, the chains have RMSD values of 0.00 with each other (i.e. A-B is 0, A-C is 0, etc).
Then with RSR, I've ensured that the NCS constraints on, and I've made sure that chain 'A' is the master and 'B-D' are the dependents in the selection panel, and I hit run. Unfortunately I'm missing something or there's an issue because I still see rotamer changes between chains. Using the Superpose Models tool in phenix I see that the RMSD between chains is slight, ~0.03 to 0.02.
What do you suppose I could be missing?
Kind Regards, Seb _______________________________________________ phenixbb mailing list -- [email protected] To unsubscribe send an email to [email protected] Unsubscribe: phenixbb-leave@%(host_name)s
Hello Pavel, As I was writing an email and repeating the steps you outlined above, the RMSDs became 0. So no Phenix level problem, but a user problem. My belief is that the NCS constraints were not toggled properly, and that there was a problem with my input model not being docked correctly. Model is now finished to my satisfaction, and all chains look the same. All good now, thank you so much for your help! -Seb
participants (2)
-
Pavel Afonine -
skaruppan＠kumc.edu