search

johnnytam100 / AlphaCutter

This repository contains the AlphaCutter.py for the removal of non-globular regions from predicted protein structures.
https://doi.org/10.1002/pmic.202300176
MIT License
13 stars 2 forks source link

AlphaCutter fails for some PDBs when using `--pLDDT_min` greater than 50 #5

Open Rohit-Satyam opened 3 days ago

Rohit-Satyam commented 3 days ago

I have realized that Alphacutter skips some PDBs when the --pLDDT_min parameters is set greater than 50. Is there a specific reason to do so?

I am testing the following PDB files from AlphaFold:

AF-Q8IFM5-F1-model_v4.pdb AF-Q8IFM8-F1-model_v4.pdb

I also see that no matter how high I set the --pLDDT_min (even 90) for some reason the poor quality domains stays See the example of the highlighted region with a low pLDDT score in Q8IFM8:

image

johnnytam100 commented 2 days ago

Hi @Rohit-Satyam

  1. increasing pLDDT cutoff didn't help because, using our default parameters (I suppose where you started with), AlphaCutter treated the N-terminal non-globular helix as a domain region AND calculated the average pLDDT as a whole (i.e. together with the nice beta structure you have in this example).

  2. Therefore, please use a smaller number for --helix_min and --loop_min and further fine-tune them

e.g. python AlphaCutter_v111.py --loop_min 5 --helix_min 10 --fragment_min 5 --domain_min 0 --pLDDT_min 0 --local_contact_range 5 --single_out --domain_out

getting

image