Added pdb_selb to filter by B-factor values

[ahmedselim2017]

Hi,

I was working on some Alphafold predictions and needed a tool to filter with B-factor and saw #163. So I have written a script called pdb_selb to filter atoms by their B-factor values.

However, as the signs < and > may interfere with shell redirect commands, I have added an option to select the operator that should be used instead of directly writing the operation and the threshold value as the same option.

Also, I have added tests and documentation for pdp_selb. But, as this is my first time contributing to this project please let me know if you have any feedback or improvement ideas!

[joaomcteixeira]

Hi @ahmedselim2017 !

Thanks for your contribution. I like the way to presented the PR, everything follows the pdb-tools strategy and architecture. I left some comments I think you should address.

@amjjbonvin @JoaoRodrigues The new tool here follows the pdb-tools "one script one job" paradigm and is something we don't have yet. Personally, I like it. Once the comments are addressed I agree with merging it. It should be [FEATURE].

Cheers,

[amjjbonvin]

PS: One more question/comment: the selection should act on a residue basis, meaning by that that full residues should be kept/removed and not only a few atoms per residue. Not an issue for pLDDT, but the B-factors are atom-specific.

[ahmedselim2017]

I have implemented a new option called filtering_mode to select if the mean (used Python's statistics.fmean instead of sum()/len() for better precision), minimum, or maximum B-factor of a residue should be used to filter residues.

While testing, I noticed that the code fails if there are nonconsecutive records for the same residue in a PDB file. The current code assumes the records of a residue should be consecutive and groups consecutive records with the same chain and residue ID to filter it.

We could mitigate this issue by sorting the PDB file using pdb_sort before filtering. Or, as this would also sort already sorted files, we could keep a list of the already processed chain and residue IDs, and if a non-processed record has the same IDs, we could throw an error and then direct the user to sort their PDB file using pdb_sort.