sirimullalab / dlscore Goto Github PK

Hi,

Thank you for releasing this code. We have been trying to implement this but found that DLScore is very slow in terms of performance, even when disabling the NNscore component, around 5 seconds per compound. Furthermore, we have not achieved good scaling when using multiple parallel instances (on the same host), observing little scale-up (<50%) when splitting the job across 10 CPUs (10 concurrent DLScore runs) and plateau around 20 CPUs.

Is this something you've observed as well and could you give some pointers as to how to improve performance?

Thanks!

Hello,

I am using DLSCORE for the first time.

The input files were obtained from a docking made using Schrodinger Suite tools (GLIDE and Induced Fit docking protocol). I am getting warnings like below (the full output is in the attached file):

....
WARNING: Duplicate receptor atom detected: "ATOM 237 N BVAL B 29 5.157 -86.693 -37.071 0.29 26.50 -0.337 N". Not loading this duplicate.
WARNING: Duplicate receptor atom detected: "ATOM 239 CA BVAL B 29 4.047 -85.753 -37.075 0.29 27.11 0.190 C". Not loading this duplicate.
WARNING: Duplicate receptor atom detected: "ATOM 241 C BVAL B 29 3.957 -85.097 -38.460 0.29 27.68 0.349 C". Not loading this duplicate.
....

Should I be concerned? Will this affect the scoring?

The protein was downloaded from Protein Data Bank and prepared (fixed) using the Protein Preparation Wizard from the Suite.

Regards,

Camps
GS1_H.txt

Hi

Thanks for providing the dlscore script.

I have been trying to run dlscore on docked files of small molecules saved in .mol2 format.
Sometimes it reads the .mol2 ligand files and sometimes the script tries to automatically search for ligands with .pdbqt by repalcing the extension: e.g:

--ligand test.mol2 # from the run command

error report :
Command-line parameters used:
Receptor: /ichec/work/nmlif042b/VS/receptor.pdbqt
Ligand: test.pdbqt
Vina executable: /ichec/work/nmlif042b/dlscore/autodock_vina_1_1_2_linux_x86/bin/vina

Traceback (most recent call last):
File "/ichec/work/nmlif042b/dlscore/dlscore.py", line 2466, in
output = ds.get_output()
File "/ichec/work/nmlif042b/dlscore/dlscore.py", line 2392, in get_output
f = open(lig,'r')
FileNotFoundError: [Errno 2] No such file or directory: 'test.pdbqt'

Thanks
Ajay

Hello,
I have just download and tried DLScore. When trying to run the test file to check that everything is ok, I got the following error:

bash test_run.sh
Using TensorFlow backend.
setting PYTHONHOME environment
setting PYTHONHOME environment
adding gasteiger charges to peptide
Command-line parameters used:
Receptor: samples/10gs/10gs_protein.pdbqt
Ligand: samples/10gs/10gs_ligand.pdbqt
Vina executable: /mnt/sda1/Shared_folder/DLSCORE-master/autodock_vina_1_1_2_linux_x86/bin/vina

Traceback (most recent call last):
File "dlscore.py", line 2467, in
output = ds.get_output()
File "dlscore.py", line 2408, in get_output
score=calculate_score(lig_array, receptor, input_parameters, self.nb_nets, temp_filename, rec, "\t")
File "dlscore.py", line 2289, in calculate_score
for dl_net in dl_nets(nb_nets):
File "dlscore.py", line 152, in dl_nets
loaded_model.load_weights(os.path.join(networks_dir, weight))
File "/mnt/sda1/Shared_folder/DLSCORE-master/.venv/lib/python3.6/site-packages/keras/engine/network.py", line 1180, in load_weights
f, self.layers, reshape=reshape)
File "/mnt/sda1/Shared_folder/DLSCORE-master/.venv/lib/python3.6/site-packages/keras/engine/saving.py", line 875, in load_weights_from_hdf5_group
original_keras_version = f.attrs['keras_version'].decode('utf8')
AttributeError: 'str' object has no attribute 'decode'

Lines 874-881 of file "saving.py" show:

:
if 'keras_version' in f.attrs:
original_keras_version = f.attrs['keras_version'].decode('utf8')
else:
original_keras_version = '1'
if 'backend' in f.attrs:
original_backend = f.attrs['backend'].decode('utf8')
else:
original_backend = None
:

but if I change both if statements to

:
if 'keras_version' in f.attrs:
original_keras_version = '1'
else:
original_keras_version = '1'
if 'backend' in f.attrs:
original_backend = None
else:
original_backend = None
:

then the calculation gets to the end giving the output:

bash test_run.sh
Using TensorFlow backend.
setting PYTHONHOME environment
setting PYTHONHOME environment
adding gasteiger charges to peptide
Command-line parameters used:
Receptor: samples/10gs/10gs_protein.pdbqt
Ligand: samples/10gs/10gs_ligand.pdbqt
Vina executable: /mnt/sda1/Shared_folder/DLSCORE-master/autodock_vina_1_1_2_linux_x86/bin/vina

DLSCORE OUTPUT: [{'vina_output': -6.7202, 'nnscore': 4.810024302277946, 'dlscore': 5.851217412948609}]

which I guess is OK.

My question: the changes made to the file saving.py might alter or affect the results of DLScore in any way or can I safely use the software in this way? I understand that the modified lines are only querying the keras and backend versions, so it might be that this is not so important for DLScore execution, is it?

Thanks for your comments.

Jordi

Hello, I've been testing out your scoring function and had a few questions. I combed through the paper but could not find where you specify the optimal number of hidden layers and number of neurons per hidden layer. Does this train a number of networks with those parameters varied and then take the top 10 best performing networks for the final score?

Another question: When I convert a PDB file with multiple ligands to a PDBQT, the code seems to only process the first ligand and not the rest, as I only get 1 score reported. Does the code require that I split the multiple ligands into separate files and process the files one at a time?

1. The "requirements.txt" file is missing which is needed in "setup.sh" file.
2. The network file can't be download (permission denied to access /files/dlscore/general.tar.gz on the server.)