Merge pull request #22 from TieuLongPhan/dev

Fix performance issues

Test/SynUtils/test_chem_utils.py

@@ -78,3 +78,8 @@ def test_remove_stereochemistry(self):
         smiles = "C[C@H][C@@H]"
         result = normalize_smiles(smiles)
         self.assertEqual("CCC", result)
+
+    def test_edge_case_1(self):
+        smiles = "F[Sb@OH12](F)(F)(F)(F)F"
+        result = normalize_smiles(smiles)
+        self.assertEqual("F[Sb](F)(F)(F)(F)F", result)

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "synrbl"
-version = "0.0.11"
+version = "0.0.12"
 authors = [
 	{name="Tieu Long Phan", email="long.tieu_phan@uni-leipzig.de"}, 
 	{name="Klaus Weinbauer", email="klaus@bioinf.uni-leipzig.de"}

synrbl/SynMCSImputer/MissingGraph/find_graph_dict.py

@@ -1,7 +1,6 @@
 import pandas as pd
 
 from rdkit import Chem
-from rdkit.rdBase import BlockLogs
 from joblib import Parallel, delayed
 from typing import List
 import multiprocessing
@@ -88,55 +87,9 @@ def find_single_graph_parallel(mcs_mol_list, sorted_reactants_mol_list, n_jobs=4
     - 'issue' (str): Any issues encountered during processing.
     """
 
-    # def process_single_pair(reactant_mol, mcs_mol):
-    #     try:
-    #         block = BlockLogs()
-    #         (
-    #             mols,
-    #             boundary_atoms_products,
-    #             nearest_neighbor_products,
-    #         ) = FindMissingGraphs.find_missing_parts_pairs(reactant_mol, mcs_mol)
-    #         del block
-    #         return {
-    #             "smiles": [
-    #                 Chem.MolToSmiles(mol) if mol is not None else None for mol in mols
-    #             ],
-    #             "boundary_atoms_products": boundary_atoms_products,
-    #             "nearest_neighbor_products": nearest_neighbor_products,
-    #             "issue": "",
-    #         }
-    #     except Exception as e:
-    #         return {
-    #             "smiles": [],
-    #             "boundary_atoms_products": [],
-    #             "nearest_neighbor_products": [],
-    #             "issue": str(e),
-    #         }
-
-    # def process_single_pair_safe(reactant_mol, mcs_mol, job_timeout=5):
-    #     pool = multiprocessing.Pool(1)
-    #     async_result = pool.apply_async(
-    #         process_single_pair,
-    #         (
-    #             reactant_mol,
-    #             mcs_mol,
-    #         ),
-    #     )
-    #     try:
-    #         return async_result.get(job_timeout)
-    #     except multiprocessing.TimeoutError:
-    #         return {
-    #             "smiles": [],
-    #             "boundary_atoms_products": [],
-    #             "nearest_neighbor_products": [],
-    #             "issue": "Find Missing Graph terminated by timeout.",
-    #         }
-    #     finally:
-    #         pool.terminate()  # Terminate the pool to release resources
     def process_single_pair(reactant_mol, mcs_mol, job_timeout=2):
         try:
-            block = BlockLogs()
-            pool = multiprocessing.Pool(1)
+            pool = multiprocessing.pool.ThreadPool(1)
             async_result = pool.apply_async(
                 FindMissingGraphs.find_missing_parts_pairs,
                 (
@@ -146,7 +99,6 @@ def process_single_pair(reactant_mol, mcs_mol, job_timeout=2):
             )
             result = async_result.get(job_timeout)
             pool.terminate()  # Terminate the pool to release resources
-            del block
             return {
                 "smiles": [
                     Chem.MolToSmiles(mol) if mol is not None else None

synrbl/SynMCSImputer/MissingGraph/find_missing_graphs.py

@@ -1,6 +1,7 @@
 import copy
 
 from rdkit import Chem
+from rdkit.rdBase import BlockLogs
 from synrbl.SynMCSImputer.SubStructure.substructure_analyzer import SubstructureAnalyzer
 from synrbl.SynMCSImputer.MissingGraph.molcurator import MoleculeCurator
 from typing import List, Dict, Optional, Tuple
@@ -74,6 +75,7 @@ def find_missing_parts_pairs(
         - list of lists: Each sublist contains the nearest neighbors of the
             corresponding molecule.
         """
+        block = BlockLogs()
         missing_parts_list = []
         boundary_atoms_lists = []
         nearest_neighbor_lists = []
@@ -198,6 +200,7 @@ def find_missing_parts_pairs(
                 boundary_atoms_lists.append(None)
                 nearest_neighbor_lists.append(None)
 
+        del block
         return missing_parts_list, boundary_atoms_lists, nearest_neighbor_lists
 
     @staticmethod

synrbl/SynMCSImputer/SubStructure/mcs_graph_detector.py

@@ -1,3 +1,5 @@
+import gc
+
 from rdkit import Chem
 from rdkit.Chem import rdFMCS
 from rdkit.Chem import rdRascalMCES
@@ -79,6 +81,7 @@ def IterativeMCSReactionPairs(
         """
 
         # Sort reactants based on the specified method
+        mcs_results = []
         if sort == "MCIS":
             if params is None:
                 params = rdFMCS.MCSParameters()
@@ -116,6 +119,9 @@ def IterativeMCSReactionPairs(
         else:
             raise ValueError("Invalid sort method. Choose 'MCS' or 'Fragments'.")
 
+        del mcs_results
+        gc.collect()
+
         mcs_list = []
         current_product = product_mol
         for reactant, _ in sorted_reactants:

synrbl/SynUtils/chem_utils.py

@@ -141,8 +141,14 @@ def remove_atom_mapping(smiles: str) -> str:
     return smiles
 
 
+def remove_stereo_chemistry(smiles: str) -> str:
+    pattern = re.compile(r"\[(?P<atom>(\w+))@+\w+\]")
+    smiles = pattern.sub(r"[\g<atom>]", smiles)
+    return smiles
+
+
 def normalize_smiles(smiles: str) -> str:
-    smiles = smiles.replace("@", "")
+    smiles = remove_stereo_chemistry(smiles)
     if ">>" in smiles:
         return ">>".join([normalize_smiles(t) for t in smiles.split(">>")])
     elif "." in smiles:
@@ -250,7 +256,6 @@ def count_radical_atoms(smiles: str, atomic_num: int) -> int:
 
     # Iterate over all atoms in the molecule
     for atom in mol.GetAtoms():
-
         if atom.GetAtomicNum() == atomic_num and atom.GetNumRadicalElectrons() > 0:
             # Further check if the atom is isolated (has no neighbors)
             if len(atom.GetNeighbors()) == 0: