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

Discngine / fpocket

fpocket is a very fast open source protein pocket detection algorithm based on Voronoi tessellation. The platform is suited for the scientific community willing to develop new scoring functions and extract pocket descriptors on a large scale level. fpocket is distributed as free open source software. If you are interested in integrating fpocket in an industrial setting and require official support, please contact Discngine (www.discngine.com).

MIT License

271 stars 60 forks source link

Only tiny pockets are found when running on 8g9b without HETATM records #134

Open pschmidtke opened 4 months ago

pschmidtke commented 4 months ago

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 commented 4 months ago

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 commented 4 months ago

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 commented 4 months ago

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 commented 4 months ago

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?