downgraded rdkit to 2020.09.5 avoid bugs

Author: Truman-Xu

.gitignore

@@ -4,4 +4,6 @@ __pycache__/
 /conda/*
 *.egg-info/*
 test_hyper.param
-
+conda-pkg/
+build/
+dist/

environment.yml

@@ -8,7 +8,12 @@ channels:
 dependencies:
   - python>=3.7
   - pytorch
-  - rdkit>=2020.03.3.0
+  - rdkit=2020.09.5
+  ## Note: The new code for Chem.MolFragmentToSmiles() will not work on the JTVAE
+  ## in the newer versions of rdkit (starting from v2021.03.1 and hopefully ends with 03.4?)
+  ## See issue https://github.com/rdkit/rdkit/issues/3998
+  ## and https://github.com/chemprop/chemprop/pull/182
+  ## The work on the patch is to be released in 2021.03.4
   - scipy
   - rxdock
   - tmap

hyper.param

@@ -2,20 +2,20 @@
 
 ############### Parameters for SnD #####################
 receptor_name = CDK2_5IEV                          # name the receptor to create dir
-receptor_file = /home/ziqiaoxu/SampleDock/targets/CDK2_5IEV/5IEV.mol2      
+receptor_file = ./targets/CDK2_5IEV/5IEV.mol2      
                                                    # must be mol2 format
-ligand_file = /home/ziqiaoxu/SampleDock/targets/CDK2_5IEV/Roniciclib.sd          
+ligand_file = ./targets/CDK2_5IEV/Roniciclib.sd          
                                                    # must be sd format
 ncycle = 1600                                      # number of cycles to be run
 ndesign = 20                                       # number of designs to be generated per cycle
-ensemble = 1                                       # top of number of design to generate the average structure 
+ensemble = 1                                       # number of top designs to generate the average structure 
                                                    # (1 being just the top scoring structure)
-seed_smi = C1(C(NC2=CC=CC=C2)=NC=N3)=C3C=CC=C1     
+nseeds = 1                                         # number of top designs to be used as seeds for distributed generation
+                                                   # nseeds overrides ensemble
+seed_smi = C1=CC=CC=C1    
                                                    # initial seeding SMILES for the first cycle, default to benzene
 
 ############### Parameters for rDock ###################
-cavity_protocol = rbcavity -was -d -r                # cmd and option for creating pocket 
-                                                   # -W option for rxdock precompiled bin file, -was for locally compiled rdock installation
 docking_prm = dock.prm                             # docking protocol (-p), no solvation term by default 
 npose = 100                                        # number of poses generated (-n)
 prefix = pose_docked_                              # prefix of the output files from rDock

sampledock/SnD/__init__.py

@@ -1,5 +1,4 @@
 from .docking import dock, sort_pose, save_pose
-from .pocket_prepare import prep_prm
 from .sampler_util import hyperparam_loader, create_wd, smiles_to_sdfile
 from .generator import single_generator, distributed_generator
 from .post_process import mkdf, combine_designs

sampledock/SnD/docking.py

@@ -13,7 +13,7 @@
 
 def dock(ligs, dock_dir, prmfile, docking_prm, npose, prefix = 'docked'):
     # ligs must be a list of file path
-    print('[INFO]: Docking in Progress\t', end = '\r')
+    print('[INFO]: Docking in Progress\t\t', end = '\r')
     sys.stdout.flush()
     procs = []
     for i,lig in enumerate(ligs):
@@ -40,7 +40,7 @@ def sort_pose(dock_dir, sort_by, prefix = None):
                       if x.endswith('.sd')]
 
     if len(poses_mols) == 0: 
-        raise Exception('No .sd file matching the criteria in %s'%dock_dir)
+        raise FileNotFoundError('No .sd file matching the criteria in %s'%dock_dir)
 
     best_poses = []
     for mol_path in poses_mols:

sampledock/SnD/pocket_prepare.py

@@ -51,10 +51,10 @@ def wrt_prm(parameter,filename='pocket_docking.prm'):
 def prep_prm(receptor,ligand,recep_name,target_dir):
 
     receptor = os.path.abspath(receptor)
-    if not os.path.exists(receptor): print(receptor+'\nRECEPTOR FILE NOT EXIST!'); return None
+    if not os.path.exists(receptor): raise FileNotFoundError(receptor+'\nRECEPTOR FILE NOT EXIST!')
 
     ligand = os.path.abspath(ligand)
-    if not os.path.exists(ligand): print(ligand+'\nLIGAND FILE NOT EXIST!'); return None
+    if not os.path.exists(ligand): raise FileNotFoundError(ligand+'\nLIGAND FILE NOT EXIST!')
 
     cav_dir = os.path.abspath(target_dir)+'/cavity'
     os.makedirs(cav_dir)
@@ -65,10 +65,16 @@ def prep_prm(receptor,ligand,recep_name,target_dir):
 
 def create_cav(prmfile):
     # rbcavity must be installed/loaded to execute the cmdline
-    cmdline = "rbcavity -was -d -r %s"%prmfile
+    cmdline = "rbcavity -W -d -r %s"%prmfile
     proc = subprocess.Popen(cmdline, shell=True)
-    proc.wait()
-    print('Docking pocket grid created for: \n'+prmfile+'\n')
+    returncode = proc.wait()
+    if returncode == 2:
+        cmdline = "rbcavity -was -d -r %s"%prmfile
+        proc = subprocess.Popen(cmdline, shell=True)
+        returncode = proc.wait()
+    if returncode == 0:
+        print('Docking pocket grid created for: \n'+prmfile+'\n')
+    else: raise Exception('Cavity creation failed')
 
 if __name__ == "__main__":
     import argparse

sampledock/__main__.py

@@ -19,7 +19,7 @@
 rdBase.DisableLog('rdApp.error')
 from .jtvae import Vocab, JTNNVAE
 # Sample and Dock tools
-from .SnD import prep_prm
+from .SnD.pocket_prepare import prep_prm, create_cav
 from .SnD import dock, sort_pose, save_pose
 from .SnD import hyperparam_loader, create_wd, smiles_to_sdfile
 from .SnD import single_generator, distributed_generator
@@ -61,10 +61,7 @@
 prmfile, cav_dir = prep_prm(p.receptor_file,p.ligand_file,p.receptor_name,wd)
 
 ## create pocket
-cmdline = p.cavity_protocol+' %s > %s/create_cavity.out'%(prmfile,cav_dir)
-proc = subprocess.Popen(cmdline, shell=True)
-proc.wait()
-print('Docking pocket grid created')
+create_cav(prmfile)
 
 ## Main loop: VAE on subsequent returned compounds
 for j in range(p.ncycle):

sampledock/jtvae/chemutils.py

@@ -67,6 +67,12 @@ def copy_edit_mol(mol):
 
 def get_clique_mol(mol, atoms):
     smiles = Chem.MolFragmentToSmiles(mol, atoms, kekuleSmiles=True)
+    ## Comment by Truman 6/23/2021
+    ## The above code will not longer work on this perticular task
+    ## with newer versions of rdkit (starting from v2021.03.1 and hopefully ends with 03.4?)
+    ## See issue https://github.com/rdkit/rdkit/issues/3998
+    ## and https://github.com/chemprop/chemprop/pull/182
+    ## The work on the patch is to be released in 2021.03.4
     new_mol = Chem.MolFromSmiles(smiles, sanitize=False)
     new_mol = copy_edit_mol(new_mol).GetMol()
     new_mol = sanitize(new_mol) #We assume this is not None
@@ -118,7 +124,9 @@ def tree_decomp(mol):
         cnei = nei_list[atom]
         bonds = [c for c in cnei if len(cliques[c]) == 2]
         rings = [c for c in cnei if len(cliques[c]) > 4]
-        if len(bonds) > 2 or (len(bonds) == 2 and len(cnei) > 2): #In general, if len(cnei) >= 3, a singleton should be added, but 1 bond + 2 ring is currently not dealt with.
+        if len(bonds) > 2 or (len(bonds) == 2 and len(cnei) > 2): 
+            # In general, if len(cnei) >= 3, a singleton should be added, 
+            # but 1 bond + 2 ring is currently not dealt with.
             cliques.append([atom])
             c2 = len(cliques) - 1
             for c1 in cnei:
@@ -134,9 +142,10 @@ def tree_decomp(mol):
                     c1,c2 = cnei[i],cnei[j]
                     inter = set(cliques[c1]) & set(cliques[c2])
                     if edges[(c1,c2)] < len(inter):
-                        edges[(c1,c2)] = len(inter) #cnei[i] < cnei[j] by construction
+                        edges[(c1,c2)] = len(inter) 
+                        #cnei[i] < cnei[j] by construction
 
-    edges = [u + (MST_MAX_WEIGHT-v,) for u,v in edges.items()] #Changed from .iteritems() to .items() by Truman for python 3.7
+    edges = [u + (MST_MAX_WEIGHT-v,) for u,v in edges.items()] 
     if len(edges) == 0:
         return cliques, edges
 

sampledock/jtvae/jtnn_vae.py

@@ -32,8 +32,6 @@ def __init__(self, vocab, hidden_size, latent_size, depthT, depthG):
         self.A_assm = nn.Linear(latent_size, hidden_size, bias=False)
         self.assm_loss = nn.CrossEntropyLoss(reduction='sum')
 
-        ## nn.Linear: Applies a linear transformation to the incoming data (y = xA^T + b)
-        ## Not sure how the biases and weights are specified this way
         self.T_mean = nn.Linear(hidden_size, latent_size)
         self.T_var = nn.Linear(hidden_size, latent_size)
         self.G_mean = nn.Linear(hidden_size, latent_size)
@@ -98,7 +96,9 @@ def find_ensemble(self,smiles_list):
             # This is due to difference in parsing of SMILES (especially rings)
             ## TODO: Convert sampledock to OOP structure and use the vectors directly
             except KeyError as key:
-                print('[KeyError]',key,'is not part of the vocabulary (the model was not trained with this scaffold)')
+                print('[KeyError]',key,\
+                'is not part of the vocabulary \
+                (the model was not trained with this scaffold)')
                 continue
             tree_mean = self.T_mean(x_tree)
             tree_log_var = -torch.abs(self.T_var(x_tree)) 
@@ -223,8 +223,11 @@ def decode(self, x_tree_vecs, x_mol_vecs, prob_decode):
 
         cur_mol = cur_mol.GetMol()
         set_atommap(cur_mol)
-        cur_mol = Chem.MolFromSmiles(Chem.MolToSmiles(cur_mol))
-        return Chem.MolToSmiles(cur_mol) if cur_mol is not None else None
+        try:
+            Chem.SanitizeMol(cur_mol)
+            return Chem.MolToSmiles(cur_mol) 
+        except rdkit.Chem.rdchem.AtomKekulizeException:
+            return None
 
     def dfs_assemble(self, y_tree_mess, x_mol_vecs, all_nodes, cur_mol, global_amap, fa_amap, cur_node, fa_node, prob_decode, check_aroma):
         fa_nid = fa_node.nid if fa_node is not None else -1

setup.py

@@ -9,7 +9,7 @@
 
 setup(
     name='sampledock',
-    version='0.5',
+    version='0.5.1',
     description='Molecular design framework the merges generative AI and molecular docking',
     author='Ziqiao Xu and Aaron Frank',
     author_email='[email protected]',