-P explicit pocket option

[lucaschestnuttree]

Hi,

I’m using fpocket to compare a central pocket with narrow tunnels across WT and mutant structures. To capture narrow necks I lower the min α-sphere radius (-m), but that tends to produce extraneous surface grooves/spheres that are continuous with the site.

(After some manual curation leaving out surface grooves, merging neighboring pockets of interest) I tried to define the site explicitly with -P (residue list), expecting that to confine the pocket, but for now my runs produce an error:

[INFO] Running: fpocket -m 3 -M 6.2 -D 3 -i 50 -v 8000 -P 107::A.121::A.209::A.212::A.213::A.216::A.221::A.244::A.247::A.300::A.303::A.304::A.307::A.308::A.312::A.389::A.393::A.500::A.501::A.502::A.503::A -f A0A087X1C5_clean.pdb
***** POCKET HUNTING BEGINS *****
! No pocket to refine! (argument NULL: (nil)).

For info I am using:

Name Version Build Channel
fpocket 4.2.2 h7b35b64_0 conda-forge

Could you also please clarify the following (implementation-level, file/function names welcome):

• Where in the algorithm -P is applied. Is it a seeding step for clustering/merging of α-spheres, or a post-filter, or both?

• Exact inclusion rules once -P is set.

• Interactions with other options. What is the defined behavior with -m, -M, -i, and merging thresholds when -P is used? Any recommended parameter ranges to capture tunnels but avoid surface grooves? When playing around with options to identify pockets by far the biggest impact for me was lowering -m.

Thanks in advance for clarifying this! And kudos on fpocket, it's by far been the handiest detection tool I've tried thus far.

Cheers,
Lucas Chataigner

[lucaschestnuttree]

Follow-up / clarification

• The initial error (! No pocket to refine!) came from a formatting mismatch in my -P string. The parser expects resSeq:iCode:chain, and when there’s no insertion code it must be written as - (e.g. 107:-:A instead of 107::A). Using :-: fixed the issue. It may be worth updating the docs to make that explicit.

• I also noticed the fparams.c file on the GitHub master branch doesn’t show the -P parsing code, even though the 4.2.2 conda-forge build does support it. That mismatch can be confusing, so pointing users to the latest tagged release might help.

• From looking at the source and with some help from an LLM, my current understanding is that -P works by controlling the seeding step: α-spheres must touch the specified residues (subject to --min-n-explicit-pocket), and then clustering expands with the normal -D connectivity rules. Could you confirm if that’s correct? It might be worth putting this description somewhere in the docs for future users.

With the correct syntax and version, -P is working as designed.

Thanks!