search

tiwarylab / GrASP

Graph Attention Site Prediction (GrASP): Identifying Druggable Binding Sites Using Graph Neural Networks with Attention
MIT License
49 stars 5 forks source link

YOU SUPPLIED TWO ATOMS 0.000000APART #11

Closed JavierSanchez-Utges closed 8 months ago

JavierSanchez-Utges commented 8 months ago

A structure crashed with this error: ERROR: THIS CODE IS KNOWN TO FAIL WHEN ATOMS ARE VIRTUALLYON TOP OF ONE ANOTHER. YOU SUPPLIED TWO ATOMS 0.000000APART. QUITTING NOW.

I think this is related to mdtraj: https://github.com/mdtraj/mdtraj/issues/1300.

I think mdtraj just clashes here: https://github.com/tiwarylab/GrASP/blob/75986d71441a589adbddffbf847ccc613eb89bbd/featurize_protein.py#L106 Even if I remove the raise e here: https://github.com/tiwarylab/GrASP/blob/75986d71441a589adbddffbf847ccc613eb89bbd/parse_files.py#L690

It still clashes.

zwsmith200 commented 8 months ago

Is this related to the ongoing altLoc discussion? If so, I would like to condense that structure prep conversation into one thread.

JavierSanchez-Utges commented 8 months ago

It seems so. I ran clean_pdb.py on the 2307 structures of my dataset and only got one atom warning:

~/miniconda3/envs/ob/lib/python3.7/site-packages/MDAnalysis/topology/guessers.py:146: UserWarning: Failed to guess the mass for the following atom types: SE
  warnings.warn("Failed to guess the mass for the following atom types: {}".format(atom_type))}

This will only work for PDB format, but seems after being cleaned with the script, atoms will be featurised without a problem.

You can now combine the three issues together if you want! Thanks!

zwsmith200 commented 8 months ago

Great, the Se atoms shouldn't be a problem because we convert Selenomethionine to Methionine later.