Only tiny pockets are found when running on 8g9b without HETATM records

[pschmidtke]

normally bigger pockets are identified in 8g9b but if one drops all HETATMS prior to default predictions only tiny ones are found everywhere.

8g9bclean.pdb.zip

Investigated:

• search_pocket & set_descriptors pass along the ligand containing pdb with the apo one (internal fpocket)
  • should not influence this behaviour

To investigate:
rpdb.r reading with hetatms vs non hetatms
clustering differences & voronoi vertices early in the process

Reproducible

[margaret-tse]

Hi Peter, I'm starting to think that this might be a general fpocket 4.0 issue, or an issue with how we installed fpocket. I'm attaching the results of running fpocket -f 8g9b.pdb with absolutely no changes to 8g9b.pdb. There are still a ton of small pockets in this case.
8g9b_no_changes.zip

[pschmidtke]

On 8g9b the original structure at least I get the ATP binding site in full for instance. Might be due to the clustering, and a bit tricky to unfiddle, but I'll give it a try

[margaret-tse]

Hi Peter, I ran 8g9b with the most recent fpocket version (master branch), and I got very different results from before. The previous version I was using was the fpocket branch tagged as version 4.0. Its not too much of an issue on my end to update our version and rerun fpocket over everything, but I do want to make sure I'm actually matching your results that show the ATP binding site first. Thanks!
8g9b.zip

[margaret-tse]

Hi @pschmidtke, side question: fpocket uses bfactors, right? In that case, how should we handle fpocket on Alphafold structures, since the bfactor is the plddt score in that case?