def exe_ml(**kwargs):
# Sort input keyword arguments
order = ['plot', 'network', 'stencil', 'layer', 'activation', 'batch_size', 'epochs', 'learning_rate', 'neg', 'angle', 'rot', 'data', 'smearing', 'hf', 'hf_correction', 'dropout', 'plotdata', 'flip', 'cut', 'dshift', 'shift', 'bias', 'interpolate', 'edge', 'edge2', 'custom_loss', 'gauss', 'unsharp_mask', 'amount', 'load_data', 'seed', 'addstring', 'stz_kappa', 'normalize', 'interpolate_lower']
kwargs = {k: kwargs[k] for k in order}
plot = kwargs.get('plot')
# Execute machine learning
'''
for job in list(itpd(*kwargs.values())):
ml(**dict(zip(kwargs.keys(), job)))
# '''
# '''
if not plot[0]: # Execute training job list with multithreading
kwargs['plotdata'] = [False]
jobs = []
job_list = list(itpd(*kwargs.values()))
for job in job_list:
# Add cpu affinity
job = job + (int(job_list.index(job)),)
jobs.append(Process(target=ml, args=job))
[j.start() for j in jobs]
[j.join() for j in jobs]
elif plot[0]:
# Execute validation job list with multithreading
for job in list(itpd(*kwargs.values())):
ml(**dict(zip(kwargs.keys(), job)))
def exe_save(**kwargs):
# Sort input keyword arguments
order = ['plot', 'network', 'stencil', 'layer', 'activation', 'batch_size', 'epochs', 'learning_rate', 'neg', 'angle', 'rot', 'data', 'smearing', 'hf', 'hf_correction', 'dropout', 'plotdata', 'flip', 'cut', 'dshift', 'shift', 'bias', 'interpolate', 'edge', 'edge2', 'custom_loss', 'gauss', 'unsharp_mask', 'amount', 'load_data', 'seed', 'addstring', 'stz_kappa', 'normalize', 'interpolate_lower']
kwargs = {k: kwargs[k] for k in order}
# Execute saving job list with multithreading
for job in list(itpd(*kwargs.values())):
save(**dict(zip(kwargs.keys(), job)))
def ModifyMolInPlace(self, reactants):
# This applies transformation for all the matching pattern found in mol
if len(self.reactantquery) != 1:
raise ReactionQueryError('ModifyMolInPlace: Only usable for',
'monomolecular transformation')
if isinstance(reactants, Chem.Mol):
reactants = [reactants]
elif isinstance(reactants, tuple):
reactants = list(reactants)
# Error test
if len(reactants) != len(self.reactantquery):
raise ReactionQueryError('Number of reactant mis-match')
for i in range(0, len(reactants)):
if not isinstance(reactants[i], Chem.Mol):
raise ReactionQueryError("Unrecognized instance'" +
type(reactants[i]) + "'")
reactants[i] = Chem.AddHs(reactants[i])
# make matrix of matching index
self.match_indexes = list()
i = 0
for reactant_name in self.reactantquery:
if isinstance(self.reactantquery[reactant_name], Chem.Mol):
match = reactants[i].GetSubstructMatches(
self.reactantquery[reactant_name])
if not match:
return tuple()
else:
self.match_indexes.append(match)
elif isinstance(self.reactantquery[reactant_name], MolQuery):
match = self.reactantquery[reactant_name].\
GetQueryMatches(reactants[i])
if not match:
return tuple()
else:
self.match_indexes.append(match)
i += 1
# make a index for combined reactants mol object
self.combined_mol_match_index = list()
for match in itpd(*self.match_indexes):
row = list()
modifier = 0
for i in range(0, len(match)):
if i != 0:
modifier = reactants[i - 1].GetNumAtoms()
row += list(
map(add, list(match[i]), [modifier] * len(match[i])))
self.combined_mol_match_index.append(row)
# Combine reactants
self.combined_mol = reactants[0]
for i in range(1, len(reactants)):
self.combined_mol = Chem.CombineMols(self.combined_mol,
reactants[i])
self.combined_mol = Chem.RWMol(self.combined_mol)
# Transform!! product_list = list()
for matches in self.combined_mol_match_index:
for transform in self.transformations:
transform(self.combined_mol, matches)
return self.combined_mol
def RunReactants(self, reactants):
if isinstance(reactants,Chem.Mol):
reactants = [reactants]
elif isinstance(reactants,tuple):
reactants = list(reactants)
# Error test
if len(reactants) != len(self.reactantquery):
raise ReactionQueryError('Number of reactant mis-match')
for i in xrange(0,len(reactants)):
if not isinstance(reactants[i],Chem.Mol):
raise ReactionQueryError("Unrecognized instance'"+type(reactants[i])+"'")
reactants[i] = Chem.AddHs(reactants[i])
# make matrix of matching index
self.match_indexes = list()
i = 0
for reactant_name in self.reactantquery:
if isinstance(self.reactantquery[reactant_name],Chem.Mol):
match = reactants[i].GetSubstructMatches(self.reactantquery[reactant_name])
if not match:
return tuple()
else:
self.match_indexes.append(match)
elif isinstance(self.reactantquery[reactant_name],MolQuery):
match = self.reactantquery[reactant_name].GetQueryMatches(reactants[i])
if not match:
return tuple()
else:
self.match_indexes.append(match)
i += 1
# make a index for combined reactants mol object
self.combined_mol_match_index = list()
for match in itpd(*self.match_indexes):
row = list()
modifier = 0
for i in xrange(0,len(match)):
if i != 0:
modifier = reactants[i-1].GetNumAtoms()
row += map(add,list(match[i]),[modifier]*len(match[i]))
self.combined_mol_match_index.append(row)
# Combine reactants
self.combined_mol = reactants[0]
for i in xrange(1,len(reactants)):
self.combined_mol = Chem.CombineMols(self.combined_mol,reactants[i])
self.combined_mol = Chem.RWMol(self.combined_mol)
# Transform!!
product_list = list()
for matches in self.combined_mol_match_index:
products = self.combined_mol.__copy__()
for transform in self.transformations:
transform(products, matches)
products = Chem.GetMolFrags(products,asMols=True,sanitizeFrags=False)
product_list.append(products)
return tuple(product_list)
def exe_save(**kwargs):
# Sort input keyword arguments
order = [
'plot', 'network', 'stencil', 'layer', 'activation', 'epochs',
'learning_rate', 'neg', 'angle', 'rot', 'data', 'smearing', 'hf',
'hf_correction', 'dropout', 'plotdata', 'addstring'
]
kwargs = {k: kwargs[k] for k in order}
# Execute saving job list with multithreading
for job in list(itpd(*kwargs.values())):
save(**dict(zip(kwargs.keys(), job)))
def exe_ml(**kwargs):
# Sort input keyword arguments
order = [
'plot', 'network', 'stencil', 'layer', 'activation', 'epochs',
'learning_rate', 'neg', 'angle', 'rot', 'data', 'smearing', 'hf',
'hf_correction', 'dropout', 'plotdata', 'addstring'
]
kwargs = {k: kwargs[k] for k in order}
# Execute machine learning
plot = kwargs.get('plot')
if not plot[0]: # Execute training job list with multithreading
kwargs['plotdata'] = [False]
jobs = []
for job in list(itpd(*kwargs.values())):
jobs.append(Process(target=ml, args=job))
[j.start() for j in jobs]
[j.join() for j in jobs]
elif plot[0]:
# Execute validation job list with multithreading
for job in list(itpd(*kwargs.values())):
ml(**dict(zip(kwargs.keys(), job)))
def exe_dg(**kwargs):
print(f'kwargs:\n{kwargs}')
# Sort input keyword arguments
order = ['N_values', 'stencils', 'ek', 'neg', 'silent', 'geometry', 'smearing', 'usenormal', 'dshift', 'interpolate', 'gauss', 'stz_kappa', 'interpolate_lower']
kwargs = {k: kwargs[k] for k in order}
# Create job list according to input arguments
job_list = list(itpd(*kwargs.values()))
if len(job_list) > 1:
# Execute job list with multithreading
jobs = []
for job in job_list:
# Add cpu affinity
job = job + (int(job_list.index(job)),)
jobs.append(Process(target=generate_data, args=job))
[j.start() for j in jobs]
[j.join() for j in jobs]
else:
# Execute job
generate_data(**dict(zip(kwargs.keys(), job_list[0])))
def exe_dg(**kwargs):
print(f'kwargs:\n{kwargs}')
# Sort input keyword arguments
order = [
'N_values', 'stencils', 'ek', 'neg', 'silent', 'geometry', 'smearing',
'usenormal'
]
kwargs = {k: kwargs[k] for k in order}
# Create job list according to input arguments
job_list = list(itpd(*kwargs.values()))
if len(job_list) > 1:
# Execute job list with multithreading
jobs = []
[
jobs.append(Process(target=generate_data, args=job))
for job in job_list
]
[j.start() for j in jobs]
[j.join() for j in jobs]
else:
# Execute job
generate_data(**dict(zip(kwargs.keys(), job_list[0])))
def GenerateRxnNet(initial_reactant, reaction_rules):
"""
Generates reaction network following the algorithm from
(Ind. Eng. Chem. Res. 2010, 49 (21), 10459-10470)
Arguments:
- initial_reactant: can be smiles or mol
- reaction_rules: can be smarts string or rdkit.Chem.rdChemReactions.
ChemicalReaction object
Return:
- list of reaction intermediates
Example:
Ethane C-H scission and C-C scission
A = generate_rxn_net('CC',['[C:1][H:2]>>[C:1].[H:2]',
'[C:1][C:2]>>[C:1].[C:2]'])
for product in A:
print(Chem.MolToSmiles(product))
Tip:
For dehydrogenation, you must number carbon and hydrogen i.e.
[C][H]>>[C].[H] (x)
[C:1][H:1]>>[C:1].[H:2] (o)
Same goes for changing bond order
[C][C]>>[C]=[C] (x)
[C:1][C:2]>>[C:1]=[C:2] (o)
Aromatization and Kekulize has some problem with several species. So, we
don't set aromatization for sanitize, and try kekulize.
TODO:
- record reactions as well (make it an option as it's expensive)
"""
# set-up reactants
if not isinstance(initial_reactant, list):
initial_reactant = [initial_reactant]
if isinstance(initial_reactant[0], str):
for i in range(0, len(initial_reactant)):
initial_reactant[i] = Chem.MolFromSmiles(initial_reactant[i],
sanitize=False)
# sanitize everything except aromatization set
_sanitize_except_aromatization(initial_reactant[i])
# Treatment necessary for radicals
# (https://github.com/rdkit/rdkit/issues/69)
for i in range(0, len(initial_reactant)):
# print Chem.MolToSmiles(initial_reactant[i])
initial_reactant[i] = Chem.AddHs(initial_reactant[i])
# sanitize everything except aromatization set
_sanitize_except_aromatization(initial_reactant[i])
# Chem.SanitizeMol(initial_reactant[i])
# Chem.Kekulize(initial_reactant[i])
for atoms in initial_reactant[i].GetAtoms():
atoms.SetNoImplicit(True)
Chem.AssignRadicals(initial_reactant[i])
# set up reactions
if not isinstance(reaction_rules, list):
reaction_rules = [reaction_rules]
if isinstance(reaction_rules[0], str):
for i in range(0, len(reaction_rules)):
try:
reaction_rules[i] = Read(reaction_rules[i])
except Exception:
reaction_rules[i] = ReactionFromSmarts(reaction_rules[i])
# generator main algorithm
unprocessed = initial_reactant
processed = []
while unprocessed:
# Pop a molecule and put it in a processed list
reactant0 = unprocessed[0]
processed.insert(0, unprocessed[0])
del unprocessed[0]
# go through all reactions
for reaction_rule in reaction_rules:
# set up reactant list.
# Generate combinatorial product list of reactants if reaction
# requires several reactants
reactant_list = itpd([list(range(1, len(processed)))],
repeat=reaction_rule.
GetNumReactantTemplates()-1)
# go through each set of reactants
for reactant_indexes in reactant_list:
# Reaction
# Make the reactant mol tuple (Runreactants only accept tuple)
reactants = (reactant0,)
for reactant_index in reactant_indexes:
reactants += (processed[reactant_index],)
# React
ele_reactions = reaction_rule.RunReactants(reactants)
# Record reactions (TODO)
# Pre-processing products
# Go through reactiosn and make a single list of products
products = []
for ele_reaction in ele_reactions:
for mol in ele_reaction:
products.append(mol)
# Treatment necessary for radicals
# (https://github.com/rdkit/rdkit/issues/69)
for mol in products:
for atoms in mol.GetAtoms():
atoms.SetNoImplicit(True)
atoms.UpdatePropertyCache(strict=False)
Chem.AssignRadicals(mol)
# Remove molecule with atoms with over valence
for i in range(len(products)-1, -1, -1):
for atoms in products[i].GetAtoms():
if PeriodicTable.GetDefaultValence(GetPeriodicTable(),
atoms.GetAtomicNum()
) < \
atoms.GetTotalValence():
del products[i]
break
# remove duplicates
# TODO. This removes also species with different charges
for i in range(len(products)-1, -1, -1):
for j in range(0, i):
if products[i].GetNumAtoms() ==\
products[j].GetNumAtoms() and \
products[i].GetNumAtoms() ==\
len(products[i].
GetSubstructMatch(products[j])):
del products[i]
break
# update unprocessed molecule list
# check for duplicate and append to unprocessed_list if missing
for mol1 in products:
inthelist = 0
for mol2 in processed:
# first check the nubmer of atoms and then
# look for substructure match
if mol1.GetNumAtoms() == mol2.GetNumAtoms() and \
mol1.GetNumAtoms() == len(mol1.GetSubstructMatch
(mol2)):
# if it's in processed list, break
inthelist = 1
break
# not in the processed list. append to unprocessed
if inthelist == 0:
unprocessed.insert(0, mol1)
# Prettify
for i in range(0, len(processed)):
# print Chem.MolToSmiles(processed[i])
processed[i] = Chem.RemoveHs(processed[i], sanitize=False)
_sanitize_except_aromatization(processed[i])
# print Chem.MolToSmiles(processed[i])
return processed
