Hi Casper,

I am doing real space refinement on a protein model build from a cryo-em map. I run the refinement with Amber gradients turned on and with secondary structure restraints (otherwise, more or less default settings).

Amber gradients should not be relevant to the issue. SS filtration is based on input model geometry before any coordinate refinement was done.

I have some issues with helix and sheet outliers.

Every time I run the refinement (after making sure that the bond distance in the helices are within the 3.5 Å)

I get a lot of bad annotation remarks, but with outliers just above 3.5 Å (see .log below):

…

removed outlier: 3.576A pdb=" N   VAL A 179 " --> pdb=" O   LEU A 175 " (cutoff:3.500A)

      Proline residue: A 180 - end of helix

      removed outlier: 3.681A pdb=" N   PHE A 188 " --> pdb=" O   LEU A 184 " (cutoff:3.500A)

      removed outlier: 3.628A pdb=" N   PHE A 189 " --> pdb=" O   ILE A 185 " (cutoff:3.500A)

      removed outlier: 3.521A pdb=" N   ILE A 190 " --> pdb=" O ALA A 186 " (cutoff:3.500A)

    Processing helix chain 'A' and resid 219 through 228

      removed outlier: 3.737A pdb=" N   MET A 227 " --> pdb=" O   GLU A 223 " (cutoff:3.500A)

…

Is this something that I just need to fix in several rounds of refinement or is there an explanation?

If you believe the distances you see in the log are wrong, I'm happy to investigate. Please send me the model file (off-list) with an explanation why do you think so. The distances (even 3.5) suggest a rather poor helix geometry since the target distance is 2.9A. If you are sure your annotations are correct, you may try to relax the threshold by setting parameter "distance_cut_n_o" to some larger value than default 3.5. Then the restraints will be imposed and hopefully the geometry of the helix will become good during refinement.

In addition, I have some bad sheet annotations such as:

…

removed outlier: 6.987A pdb=" N   VAL B 433 " --> pdb=" O   LEU B 440 " (cutoff:3.500A)

      removed outlier: 4.659A pdb=" N   VAL B 442 " --> pdb=" O   LEU B 431 " (cutoff:3.500A)

      removed outlier: 6.296A pdb=" N   LEU B 431 " --> pdb=" O   VAL B 442 " (cutoff:3.500A)

      removed outlier: 5.062A pdb=" N   ALA B 444 " --> pdb=" O   THR B 429 " (cutoff:3.500A)

      removed outlier: 6.129A pdb=" N   THR B 429 " --> pdb=" O   ALA B 444 " (cutoff:3.500A)

…

When I look at the model, the residues are definitely a part of the sheet – the long distance between the N-O atoms is due to the O being “flipped” (hope it makes sense).

Is this because of wrong annotation?

If you believe the residues are part of the sheet - then the annotation is correct, the model is wrong. If you believe the O needs to be "flipped" - go ahead and flip it. Do any other adjustments you deem necessary to improve the model and hydrogen-bonding pattern. This may result in proper distance between N and O and established restraint. Such big model changes are better done manually rather than during refinement with restraints.

Some relevant tutorials are:

https://www.youtube.com/watch?v=9dCkAdR1RDk&feature=youtu.be

https://www.youtube.com/watch?v=qB8W_6yuw5k&feature=youtu.be

Relevant documentation and links in it:

https://www.phenix-online.org/documentation/reference/secondary_structure.html

Please let us know if you have more questions.

Best regards,

Oleg Sobolev.

Best

Casper

_______________________________________________
phenixbb mailing list
phenixbb@phenix-online.org
http://phenix-online.org/mailman/listinfo/phenixbb
Unsubscribe: phenixbb-leave@phenix-online.org