Python-based
Hierarchical
ENvironment
for
Integrated
Xtallography
Refinement may be stabilized by addition of hydrogen-bond restraints to maintain (or idealize) secondary structure. This is particularly important at low resolution, where the protein geometry should be kept as precise as possible to avoid overfitting, but it appears to help even at moderate resolutions (up to 2.0 A, maybe further). Applying these restraints usually improves R-free by up to 0.005, and often keeps R-work and R-free from diverging as much as they might otherwise. (Note that this is not always true, and depends strongly on the weighting and outlier filtering; for this reason, we recommend experimenting with this setting to see whether and how it helps your structure, rather than blindly enabling this feature.)
To activate secondary structure restraints in phenix.refine, simply add the parameter secondary_structure.enabled=True on the command line, or check the box labeled "Secondary structure restraints" in the configuration window of the phenix.refine GUI. Assignment of atom selections will be fully automatic. We recommend using explicit hydrogens in refinement, but hydrogen bonds will be specified between heavy atoms anyway.
Hydrogen bond lengths for protein secondary structure are restrained at 2.9A with sigma 0.05A for N-O. Outliers (defined as N-O bonds greater than 3.5A) will be automatically removed, but you may prevent this with secondary_structure.protein.remove_outliers=False. This may be helpful for very poor low-resolution structures, e.g. to force proper structure on a manually built helix, but it will also cause completely spurious bond restraints to be overlooked.
Annotation of secondary structure is performed by KSDSSP, an open-source implementation by UCSF's Computer Graphics Laboratory (authors of the molecular graphics program Chimera) of the original DSSP algorithm (Kabsch and Sander 1983). Users may also provide their own annotation in the form of HELIX and SHEET records from the PDB; however, because this format is difficult to use, we recommend using the configuration syntax used in phenix.refine, and/or visually editing the secondary structure in the PHENIX GUI (both described below).
The PDB format supports 10 different types of helix, but only three of these (alpha, pi, and 3_10) are common in naturally occurring proteins and have easily deciphered hydrogen-bonding rules. User-specified helices default to alpha form (hydrogen bonds between residue n and residue n+4). Sheets may be either parallel or antiparallel. Because additional details are required to establish the hydrogen bonding pattern, we do not recommend defining sheets manually unless you are certain you know what to do. Future versions will include a more intelligent editor to assist in annotation.
Hydrogen bonding and stacking interactions stablilize structure of nucleic acids. Restraints include base-pairing restraints (hydrogen bonding and keeping basepairs in plane) and stacking restraints to keep stacked bases parallel to each other. These restraints are turned on with the main switch secondary_structure.enabled=True. They could be turned off individually: secondary_structure.nucleic_acid.enabled=False. Annotation is performed automatically or could be supplied in parameter file.
Documentation coming soon!
- Only bond length is restrained in proteins, angles may move freely.
- The distribution of oxygen-nitrogen distances in beta sheets appears to be very slightly bimodal, probably due to the difference in geometry between parallel and antiparallel sheets. Increasing the slack may compensate for this, but it isn't clear whether this is necessary.
- Although hydrogen bond outliers are filtered by default, very little validation of secondary structure definitions is performed.
- Kabsch, W., & Sander, C. (1983). Dictionary of Protein Secondary Structure: Pattern Recognition of Hydrogen-Bonded and Geometrical Features. Biopolymers, 22, 2577-2637.
- Fabiola et al. (...)