Rejection criteria

Lists and WHY_NOT

The main purpose of the Lists section of the PDB facilities is to provide rapid access to computationally derived protein structure data to protein structure bioinformaticians. A large number of Lists-entries (like entries that do not hold any protein...) would make no sense. Further, many files cannot be produced at all. The WHY_NOT system explains in all these cases why Lists-entries don′t exist. Some very general WHY_NOT information is available in the WHY_NOT system. Additionally, if files cannot be made, like ligand contacts for a file without ligands, secondary structure for a file that holds DNA only, cis prolines for a file without prolines, etc., there will be a Lists-entry called xxxx.yyy.whynot, in which xxxx is the PDB identifier and yyy is the three-letter code of the List. So, the absense of ions in 1crn leads to a WHY_NOT file with the filename 1crn.ion.whynot:

COMMENT: No metal ions found
WHATIF_PDB_ion,1crn

PDB file acceptance/rejection criteria

Files can be rejected for a large number of reasons, some of which are trivial, like the PDB file for Vancomycin that holds just a funny molecule, but no protein or nucleic acid. A few hundred files have been manually rejected over the years because of a variety of reasons like, the structure seems fully incorrect, the structure was determined under extreme conditions, the file contains just a small peptide, the structure contains so many ligands or ions that the protein fold is most likely non-physiological, the structure has been solved using some experimental refinement technique, or the structure was solved at a very low resolution. A few files are rejected because they contain problems WHAT IF cannot cope with. The list of manually rejected PDB entries is available.

Even when a file passes these filters, it is not guaranteed that it makes it through. Files can be rejected on the fly for many reasons:

• Viral and ribosomal proteins tend to live in highly packed environments and the PDB file coordinates thus, generally, are missing a lot of packing environment. Furthermore, viral and risosomal proteins generally contribute only little to protein structure related bioinformatics studies. So, viral and ribosomal proteins are missing in most Lists directories.
• Structures solved by NMR normally are solved with lower accuracy and lower precision than structures solved by X-ray. Additionally, those NMR ensembles are a pain for protein structure bioinformaticians to work with. So, structures solved by NMR are missing in most Lists directories.
• PDB files that hold more than 5000 entities (amino acids, nucleic acids, sugar units, ligands, ions, or groups of waters) or more than 50000 atoms are skipped, and so are files that contain fewer than 25 entities and files that hold fewer than 5 amino acids.
• Many PDB files hold residues for which only the Cα is present. If the number of Cα-only residues is greater than 20 or higher than 10% of the total number of residues, the PDB file is skipped.
• Residues that miss certain atoms are called incorrect. There are also other reasons why residues can be incorrect. If the number of correct residues is less than 20, the PDB file is skipped. If the number of incorrect residues is larger than 20 or larger than 5% of the total number of residues, the PDB file gets skipped too.
• Some PDB files were refined using multiple models to better match the data. Albeit that this is an interesting concept, the resulting files are rather useless for most protein structure bioinformatics projects, so any file that holds MODEL cards is skipped.

Completeness

Each file that is not rejected because of manual rejection (as listed in this file), is indicated by a *.whynot file. So for each PDB file that enters the Lists software either a results file comes out, or a *.whynot file.

We maintain an internal system that checks the whole Lists directory tree for completeness. Honesty dictates that we should tell you that if somewhere a handful of files is missing, we will not address that problem till the next quarterly complete overhaul.

In March 2018 the following textx were possible for *.WHYNOT files:

• Not enough residues
• Not enough intact residues
• Bioinformatics unfriendly aspects
• Not enough amino acids
• Not enough intact amino acids
• Too many bad residues
• Percentage bad residues too high
• MODEL records found
• Not an X-ray structure
• Too many C-alpha-only residues
• Percentage C-alpha-only residues too high
• No metal ions found
• No short contacts to metal ions
• No salt bridges found
• Too many residues
• Virus or ribosome
• Homology model
• Contains no cysteines
• Contains no cysteine bridges
• Contains no ligands
• Contains no nucleic acids

If at some time, we decide to add a few more of these WHYNOT reasons, it is likely that we forget to update this part of the documentation... But you get the drift...