Error preprocessing protein-ligand complex

[jbderoo]

Hi all,

First and foremost, thank you very much for making all your hard work publicly accessible. It is a wonderful tool tackling a challenging problem.

I am having troubles running the preprocss_pdb.sh script. I can successfully run it on the example! But I cannot run it on my own inputs with the ligand included; it does work without it. Just for clarity:

Works:
./preprocess_pdb.sh example/1a7x.pdb 1A7X_A -l FKA_B -s example/1a7x_C_FKA.sdf -o example/output/

Works:
./preprocess_pdb.sh example/1a7x.pdb 1A7X_A -o example/output/

Works:
./preprocess_pdb.sh /inputs/8G9P.pdb 8G9P_D -o example/output/

DOES NOT work:
./preprocess_pdb.sh /inputs/8G9P.pdb 8G9P_D -l YV2_D -s /inputs/YV2.sdf -o example/output/

And has error:

Precomputing features on /inputs/8G9P.pdb
Including ligand YV2 in the surface.
Mismatch between PDB and SDF ligands. Determining bond types with OpenBabel...
[00:10:22] WARNING: More than one matching pattern found - picking one
Fixing mesh...
Removing degenerated triangles
Fixmesh done!
Computing APBS...
### PDB2PQR ###
 INFO:PDB2PQR v3.5.2: biomolecular structure conversion software.
INFO:Please cite:  Jurrus E, et al.  Improvements to the APBS biomolecular solvation software suite.  Protein Sci 27 112-128 (2018).
INFO:Please cite:  Dolinsky TJ, et al.  PDB2PQR: expanding and upgrading automated preparation of biomolecular structures for molecular simulations. Nucleic Acids Res 35 W522-W525 (2007).
INFO:Checking and transforming input arguments.
INFO:Loading topology files.
INFO:Loading molecule: 8G9P_D.pdb
INFO:Setting up molecule.

### APBS ###
 Vio_ctor2: some error occurred.
Vio_ctor: Vio_ctor2() failed.
Error while opening input file 8G9P_D.in!

### MULTIVALUE ###
 Vio_ctor2: some error occurred.
Vio_ctor: Vio_ctor2() failed.
Vgrid_readDX: Problem opening virtual socket 8G9P_D.dx
main:  Problem reading standard OpenDX-format grid from 8G9P_D.dx

Traceback (most recent call last):
  File "/masif-neosurf/masif/source//data_preparation/01-pdb_extract_and_triangulate.py", line 125, in <module>
    vertex_charges = computeAPBS(regular_mesh.vertices, out_filename1+".pdb", out_filename1, mol2_file)
  File "/masif-neosurf/masif/source/triangulation/computeAPBS.py", line 64, in computeAPBS
    chargefile = open(tmp_file_base + "_out.csv")
FileNotFoundError: [Errno 2] No such file or directory: '/masif-neosurf/example/output/tmp//8G9P_D_out.csv'
/masif-neosurf

8G9P is a challenging protein-protein-ligand interface that I was keen to test via MaSIF-neosurf. Unfortunately I cannot seem to get passed the technical hurdles by myself - I tried to dive into the APBS but I am admittedly very weak with regards to using this package. Any guidance or tweaks you could provide in overcoming this would be immensely helpful. I have tried this with a few other similar complexes and encounter the same errors.

The SDF file of the ligand was generated by extracting out the small molecule's atom coords and connectivity, then converting the PDB ligand file to SDF ligand via OpenBabel (specifically, obabel YV2.pdb -O YV2.sdf). I loaded the SDF file into PyMOL along with 8G9P to ensure the molecules match identically, and they do with the exception of a few extra double bonds added in linear sections of the ligand (inspired by the Mismatch between PDB and SDF ligands. Determining bond types with OpenBabel... warning). I've included all files relevant to my work here for you to mirror my error (hopefully, anyway).

I look forward to hearing from you regarding this issue.

Kind regards,
Jacob

masif_neo_issue.tar.gz

[rukibuki]

Not a masif developer or anything but for what it is worth, I dl your files and ran your command and it worked fine from my side.

here are some of the log output:

APBS

MULTIVALUE

APBS done!
8G9P_D
Reading data from input ply surface files.
Dijkstra took 1.66s
Only MDS time: 7.92s
Full loop time: 11.27s
MDS took 11.27s
/usr/local/lib/python3.6/site-packages/numpy/core/_asarray.py:136: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
return array(a, dtype, copy=False, order=order, subok=True)
8G9P_D
Reading data from input ply surface files.
Dijkstra took 3.07s
Only MDS time: 11.86s
Full loop time: 15.66s
MDS took 15.66s
/usr/local/lib/python3.6/site-packages/numpy/core/_asarray.py:136: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
return array(a, dtype, copy=False, order=order, subok=True)
Running masif site on /home/masif-neosurf/input/8G9P.pdb
Setting model_dir to nn_models/all_feat_3l/model_data/
Setting feat_mask to [1.0, 1.0, 1.0, 1.0, 1.0]
Setting n_conv_layers to 3
Setting out_pred_dir to output/all_feat_3l/pred_data/
Setting out_surf_dir to output/all_feat_3l/pred_surfaces/
(12, 2)
WARNING:tensorflow:From /home/masif-neosurf/masif/source/masif_modules/MaSIF_site.py:108: calling reduce_sum (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version.
Instructions for updating:
keep_dims is deprecated, use keepdims instead
global_desc shape:
2025-01-30 13:01:44.136622: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
<tf.Variable 'mu_rho_0:0' shape=(1, 12) dtype=float32_ref>
12
<tf.Variable 'mu_theta_0:0' shape=(1, 12) dtype=float32_ref>
12
<tf.Variable 'sigma_rho_0:0' shape=(1, 12) dtype=float32_ref>
12
<tf.Variable 'sigma_theta_0:0' shape=(1, 12) dtype=float32_ref>

and here is an image of the ply model:

[jbderoo]

Well that is very unfortunate and frustrating - I'm using their Dockerfile! Thank you for your comment, I'll start fresh I suppose.

[jbderoo]

Hey @rukibuki, would you mind sharing your environment? Did you use their Dockerfile? Did you build the environment yourself, make any changes to their default specs listed in either their Dockerfile or their ReadMe? Thank you for your help!