def _similarity( self, id0, id1, **kwarg ) :
# Retrieves the parent molecules of the MCS.
mol0 = KBASE.ask( id0 )
mol1 = KBASE.ask( id1 )
if (mol0.total_charge() != mol1.total_charge()) :
return 0.0
return 1.0
def create(basic_graph, mcs_ids, rule, add_attr=True):
"""
Returns a graph. Node = molecule's ID or name, edge = similarity score
@type mcs_ids : C{list} of C{str}
@param mcs_ids : A list of common substructures' IDs
@type rule : C{Rule}
@param rule : The rule to determine the similarity score between two structures
"""
g = copy.deepcopy(basic_graph)
for id in mcs_ids:
id0, id1 = mcs.get_parent_ids(id)
simi = rule.similarity(id0, id1, mcs_id=id)
if (simi > 0):
if (add_attr):
try:
partial_ring = int(KBASE.ask(id, "partial_ring"))
except LookupError:
partial_ring = 0
try:
slack_simi = KBASE.ask(id, "slack_similarity")
except LookupError:
slack_simi = 0.0
g.add_edge(id0,
id1,
similarity=simi,
slack_similarity=slack_simi,
partial_ring=partial_ring,
mcs_id=id)
else:
g.add_edge(id0, id1, similarity=simi)
return g
def _similarity( self, id0, id1, **kwarg ) :
# Uses the first common substructure.
num_atom_mcs = KBASE.ask( kwarg["mcs_id"], "num_heavy_atoms" )
num_atom_mol0 = len( KBASE.ask( id0 ).heavy_atoms() )
num_atom_mol1 = len( KBASE.ask( id1 ).heavy_atoms() )
return float( (num_atom_mcs >= self._threshold ) or
(num_atom_mol0 < self._threshold + 3) or
(num_atom_mol1 < self._threshold + 3) )
def annotate_nodes_with_smiles(g):
"""
"""
for molid in g.nodes():
try:
smiles = KBASE.ask(molid, "SMILES")
except LookupError:
smiles = KBASE.ask(molid).smiles()
KBASE.deposit_extra(molid, "SMILES", smiles)
g.node[molid]["SMILES"] = smiles
def read_n_files( filenames ) :
"""
`filenames' is a list of file names. Reads the files and deposits them into the `KBASE'. Returns a list of keys.
"""
strucid = []
for fn in filenames :
strucs = read_file( fn )
for e in strucs:
id = KBASE.deposit( e.id(), e )
KBASE.deposit_extra(id, "filename", (fn))
e.set_id( id )
strucid.append( id )
return strucid
def read_n_files( filenames ) :
"""
`filenames' is a list of file names. The format of each file will be determined from the file's extension name. Reads
the files and deposits them into the `KBASE'. Returns a list of keys.
"""
strucid = []
for fn in filenames :
strucs = read_file( fn )
for e in strucs :
id = KBASE.deposit( e.id(), e )
KBASE.deposit_extra(id, "filename", (fn))
e.set_id( id )
strucid.append( id )
return strucid
def annotate_edges_with_smiles(g):
"""
"""
for e in g.edges(data=True):
try:
mcs_id = e[2]["mcs_id"]
try:
smiles = KBASE.ask(mcs_id, "SMILES")
except LookupError:
smiles = mcs.get_struc(mcs_id).smiles()
KBASE.deposit_extra(mcs_id, "SMILES", smiles)
g[e[0]][e[1]]["SMILES"] = smiles
except KeyError:
pass
def annotate_nodes_with_title(g):
"""
"""
for molid in g.nodes():
g.node[molid]["title"] = KBASE.ask(molid).title()
g.node[molid]["label"] = molid[:7]
def add_mcs_id( mcs_id_list, graph ):
for edge in graph.edges(data = True):
mol0_id = edge[0]
mol1_id = edge[1]
mol0 = KBASE.ask( mol0_id )
mol1 = KBASE.ask( mol1_id )
name0 = mol0.title()
name1 = mol1.title()
mcs_title = "mcs@%s..%s" % (name0, name1,)
mcs_id = hashlib.sha1( mcs_title ).hexdigest()
if mcs_id not in mcs_id_list:
# the first mcs_id is not in id list need to generate reverse one
mcs_title = "mcs@%s..%s" % (name1, name0,)
mcs_id = hashlib.sha1( mcs_title ).hexdigest()
if mcs_id not in mcs_id_list:
sys.exit()
graph.add_edge(mol0_id, mol1_id, mcs_id = mcs_id)
def get_parent_ids(mcs_id):
"""
Returns a pair of IDs of the common substructure's parents.
@type mcs_id: C{str}
@param mcs_id: ID of the common substructure
"""
return KBASE.ask(mcs_id, "mcs-parents")
def similarity( self, id0, id1, **kwarg ) :
try :
mcs_id = kwarg["mcs_id"]
simi = KBASE.ask( mcs_id, "similarity" )
except KeyError :
simi = Rule.similarity( self, id0, id1, **kwarg )
if (simi < self._cutoff) :
simi = 0.0
return simi
def annotate_edges_with_matches(g):
"""
"""
for e in g.edges(data=True):
try:
mcs_id = e[2]["mcs_id"]
mol0 = KBASE.ask(e[0])
mol1 = KBASE.ask(e[1])
mcs_matches = KBASE.ask(mcs_id, "mcs-matches")
trimmed_mcs = KBASE.ask(mcs_id, "trimmed-mcs")
layout_mcs = KBASE.ask(mcs_id, "layout_mcs")
g[e[0]][e[1]]["original-mcs"] = {
e[0]: mol0.smarts(mcs_matches[e[0]]),
e[1]: mol1.smarts(mcs_matches[e[1]]),
}
g[e[0]][e[1]]["trimmed-mcs"] = trimmed_mcs
g[e[0]][e[1]]["layout_mcs"] = layout_mcs
except KeyError:
pass
def _similarity( self, id0, id1, **kwarg ) :
# Uses the first common substructure.
mcs_id = kwarg["mcs_id"]
mcs0 = mcs.get_struc( mcs_id )
mol0 = KBASE.ask( id0 )
mol1 = KBASE.ask( id1 )
num_heavy_atoms = len( mcs0.heavy_atoms() )
num_light_atoms = len( mcs0.atom ) - num_heavy_atoms
KBASE.deposit_extra( mcs_id, "num_heavy_atoms", num_heavy_atoms )
KBASE.deposit_extra( mcs_id, "num_light_atoms", num_light_atoms )
return similarity.by_heavy_atom_count( mol0, mol1, mcs0 )
def matrix ( mols, mcs_ids, rule ):
import numpy
id_list = []
id_vs_simi = {}
id_vs_title = {}
title_vs_simi = {}
title_list = []
filename_vs_title = {}
for mol in mols:
#generate dictionary of id vs title of giving mols
title = mol.title()
id = mol.id()
file_path = KBASE.ask (id, "filename")
filename = os.path.basename(file_path)
if id not in id_list:
id_list.append(id)
if title not in title_list:
title_list.append(title)
id_vs_title [id] = title
filename_vs_title [filename] = title
for id in mcs_ids:
#generate dictionary of pair's title vs similarity score
id0, id1 = mcs.get_parent_ids(id)
simi = rule.similarity( id0, id1, mcs_id = id )
title0 = id_vs_title[id0]
title1 = id_vs_title[id1]
title_vs_simi [(title0,title1)] = simi
#generate the score matrix
size = len( title_list )
scores = numpy.zeros( (size, size,) )
for i in range( size ) :
scores[i, i] = 1.0
for j in range( i + 1, size ) :
title_i = title_list[i]
title_j = title_list[j]
if title_vs_simi.has_key((title_i,title_j)):
simi = title_vs_simi[(title_i,title_j)]
scores[i, j] = simi
scores[j, i] = simi
return (title_list, id_list,filename_vs_title, scores)
def get_struc(mcs_id):
"""
get the mcs strcuture based on mcs_id
"""
title = KBASE.ask(mcs_id)
id0, id1 = KBASE.ask(mcs_id, "mcs-parents")
mcs_matches = KBASE.ask(mcs_id, "mcs-matches")
atom_match0, atom_match1 = mcs_matches[id0], mcs_matches[id1]
mol0 = KBASE.ask(id0)
mol1 = KBASE.ask(id1)
mcs = KBASE.ask(id0).extract(atom_match0)
for i, e in enumerate(atom_match1, start=1):
mcs.atom_prop[i]["mapped_index"] = e
mcs.set_title(title)
mcs.set_id(mcs_id)
return mcs
def deposit_to_kbase(id0, id1, atom_match0, atom_match1):
"""
Deposits a MCS substructure and relevant information into the kbase and returns its ID in the C{KBASE}.
@type id0: C{str}
@param id0: ID of the first (reference) molecule in the C{KBASE}
@type id1: C{str}
@param id1: ID of the second molecule in the C{KBASE}
@type atom_match: C{list} of C{int}
@param atom_match: A list of atom indices of matches atoms in the reference molecule
@type mcs_mol: C{Struc}
@param mcs_mol: C{Struc} object of the MCS substructure
"""
mol0 = KBASE.ask(id0)
mol1 = KBASE.ask(id1)
name0 = mol0.title()
name1 = mol1.title()
mcs_title = "mcs@%s..%s" % (
name0,
name1,
)
mcs_id = hashlib.sha1(mcs_title).hexdigest()
mcs_id = KBASE.deposit(mcs_id, mcs_title)
# Sorts the two lists according to the ascending order of atom indices of the first list.
atom_match0, atom_match1 = zip(*sorted(zip(atom_match0, atom_match1),
cmp=lambda x, y: x[0] - y[0]))
atom_match0, atom_match1 = list(atom_match0), list(atom_match1)
KBASE.deposit_extra(mcs_id, "mcs-parents", (
id0,
id1,
))
KBASE.deposit_extra(mcs_id, "mcs-matches", {
id0: atom_match0,
id1: atom_match1,
})
return mcs_id
def main(molid_list, opt, args):
"""
@type molid_list: C{list} of C{str}'s
@param molid_list: A list of molecule IDs in the C{KBASE}
"""
#load mols files
if (opt.graph):
g = pickle.load(open(opt.graph))
else:
mols = []
for id in molid_list[opt.receptor:]:
mols.append(KBASE.ask(id))
#choose mcs search engine and rules
if (struc.infrastructure == "schrodinger"):
mcs_engine = mcs.SchrodMcs(1)
basic_rule = rule.Mcs(
rule.EqualCharge(),
rule.TrimMcs(True, rule.MinimumNumberOfAtom()))
slack_rule = rule.Mcs(
rule.EqualCharge(),
rule.TrimMcs(False, rule.MinimumNumberOfAtom()))
elif (struc.infrastructure == "oechem"):
mcs_engine = mcs.OeMcs()
basic_rule = rule.Mcs(
rule.EqualCharge(),
rule.TrimMcs_oe(True, rule.MinimumNumberOfAtom()))
slack_rule = rule.Mcs(
rule.EqualCharge(),
rule.TrimMcs_oe(False, rule.MinimumNumberOfAtom()))
logging.info("MCS searching...")
mcs_ids = mcs_engine.search_all(mols, opt)
logging.info("MCS searching... Done")
#build score matrix from mcs search enable Jonathan's graph planning algorithm
if (opt.build):
import build
(title_list, id_list, filename_vs_title,
strict_score) = build.matrix(mols, mcs_ids, basic_rule)
(title_list, id_list, filename_vs_title,
unstrict_score) = build.matrix(mols, mcs_ids, slack_rule)
import GraphGenerator4 as gg4
knownCompoundsList = []
#load the name list of coumpounds with known experimental value if there is any
try:
with open(args[0] + "/knownCompounds") as kcFile:
knownCompoundsList = kcFile.readlines()
knownCompoundsList = [
filename_vs_title[name.strip()]
for name in knownCompoundsList
]
except IOError:
print "No Known Compounds Listed"
gg = gg4.GraphGenerator4(strict_score, unstrict_score, 0.05, 6,
title_list, id_list, knownCompoundsList)
g = gg.getGraphObject()
build.add_mcs_id(mcs_ids, g)
c = networkx.connected_component_subgraphs(g)
# Gets graph (`g') and clusters (`c') using schrodinger's graph planning algorithm
else:
logging.info("Creating graph...")
g, c = graph.gen_graph(mcs_ids,
basic_rule,
slack_rule,
simi_cutoff=0.05,
max_csize=100,
num_c2c=1)
graph.annotate_nodes_with_smiles(g)
graph.annotate_nodes_with_title(g)
graph.annotate_edges_with_smiles(g)
graph.annotate_edges_with_hexcode(g)
graph.annotate_edges_with_matches(g)
logging.info("Creating graph... Done")
logging.debug(
"DEBUG: %d clusters (counted as the connected components in the graph):"
% len(c))
c.sort(lambda x, y: len(x) - len(y))
for i, e in enumerate(c):
logging.debug("DEBUG: cluster #%d, %d structures:" % (
i,
len(e),
))
titles = [KBASE.ask(id).title() for id in e]
titles.sort()
for t in titles:
logging.debug("DEBUG: %s" % t)
# store graph for reusing and analysing
pkl_fname = opt.output + ".pkl"
pkl_fh = open(pkl_fname, "w")
pickle.dump(g, pkl_fh)
pkl_fh.close()
try:
#use pygraphviz for graph layout
import graphviz
ag = networkx.to_agraph(g)
ag.node_attr["fixedsize"] = True
ag.edge_attr["penwidth"] = 2.0
simi = [float(e.attr["similarity"]) for e in ag.edges()]
scale = 1.0 / max(simi)
for e in ag.edges_iter():
try:
partial_ring = int(e.attr["partial_ring"])
except (ValueError, TypeError):
partial_ring = 0
saturation = float(e.attr["similarity"]) * scale
saturation = 0.0 if (saturation < 0) else (1.0 if (
saturation > 1) else saturation)
e.attr["color"] = "0.8396,%f,0.8" % saturation
e.attr["weight"] = saturation
del e.attr["label"]
if (saturation < 0.01 or partial_ring):
e.attr["style"] = "dashed"
ag.write(opt.output + ".dot")
except ImportError:
logging.warn(
"WARNING: pygraphviz is not installed. Cannot write a .dot output file."
)
edges = g.edges(data=True)
logging.info("%d edges in total" % len(edges))
#generate schrodinger FEP input files
if (opt.siminp):
if (opt.siminp_type == "gro"):
raise NotImplementedError(
"Support for writing Gromacs input files is not yet implemented."
)
if (opt.siminp_type == "mae"):
import schrodinger.application.desmond.fep_mapping as dfm
tmp_mae_fname = mcs.tempfile_basename + "_siminp.mae"
receptor_mol = []
if (opt.receptor):
for e in range(opt.receptor):
mol = KBASE.ask(molid_list[e])
mol._struc.property["s_leadoptmap_moltype"] = "receptor"
receptor_mol.append(mol)
for id0, id1, attr in edges:
mol0 = KBASE.ask(id0)
mol1 = KBASE.ask(id1)
out_fname = "%s_%s_%s.mae" % (
opt.siminp,
id0[:7],
id1[:7],
)
mol0._struc.property["s_leadoptmap_moltype"] = "ligand"
mol1._struc.property["s_leadoptmap_moltype"] = "%s:%s" % (
id0,
id1,
)
mol0.write(tmp_mae_fname, mode="w")
mol1.write(tmp_mae_fname, mode="a")
try:
overwrite = True
data = dfm.get_atom_mapping_data(tmp_mae_fname, atomtype=3)
if (opt.receptor):
overwrite = False
receptor_mol[0].write(out_fname, mode="w")
for i in range(1, opt.receptor):
receptor_mol[i].write(out_fname, mode="a")
dfm.write_fepsubst_to_file(data,
out_fname,
overwrite=overwrite)
except (
RuntimeError,
NameError,
):
logging.warn(
"WARNING: Failed to write the input files for '%s' and '%s'."
% (
mol0,
mol1,
))
if (not opt.save):
tmp_fnames = glob.glob(mcs.tempfile_basename + "*")
for fname in tmp_fnames:
os.remove(fname)
for e in strucs:
id = KBASE.deposit( e.id(), e )
KBASE.deposit_extra(id, "filename", (fn))
e.set_id( id )
strucid.append( id )
return strucid
infrastructure = "oechem"
except ImportError, e :
pass
if (infrastructure is None) :
print "ERROR: Need either Schrodinger's or OEChem's infrastructure to run, but none is found."
import sys
sys.exit( 1 )
if ("__main__" == __name__) :
filenames = ["xfer3.10.mol2", "xfer3.11.mol2",]
id_list = read_n_files( filenames )
mol0 = KBASE.ask( id_list[0] )
print mol0.title(), len( mol0.heavy_atoms() )
mol1 = KBASE.ask( id_list[1] )
print mol1.title(), len( mol1.heavy_atoms() )
def search_all(self, mols, opt):
if (not opt.mcs):
mae_fname = tempfile_basename + ".mae"
out_fname = tempfile_basename + ".csv"
log_fname = tempfile_basename + ".log"
log_fh = open(log_fname, "w")
if (os.path.isfile(mae_fname)):
os.remove(mae_fname)
for mol in mols:
title = mol.title()
mol.set_title(mol.id())
mol.write(mae_fname)
mol.set_title(title)
cmd = [
self._cmd,
"-imae",
mae_fname,
"-opw",
out_fname,
"-atomtype",
str(self._typing),
"-nobreakring",
]
mcs_proc = subprocess.Popen(cmd,
stderr=subprocess.STDOUT,
stdout=log_fh)
null, stderr = mcs_proc.communicate()
val = mcs_proc.returncode
if (val == 17):
raise RuntimeError(
"Used a MCS feature that requires Schrodinger's CANVAS_ELEMENTS license."
)
if (val != 0):
msg = "CanvasMCS exited prematurely. This could be because the input molecules were too dissimilar" \
" or too numerous, or because the chosen atom-typing scheme was too general."
with open(out_fname) as fh:
msg += "\n\n"
msg += fh.read()
raise RuntimeError(msg)
else:
logging.debug(
"DEBUG: Reuse previous MCS searching results: '%s'." %
opt.mcs)
out_fname = opt.mcs
with open(out_fname, "r") as fh:
import csv
lines = fh.readlines()[1:]
mcs_match = []
for tokens in csv.reader(lines):
mcs_match.append(
McsMatch(tokens[1], tokens[3], tokens[11], tokens[14],
tokens[9], tokens[12]))
ret = []
for m in mcs_match:
id0 = m.mol0_id
id1 = m.mol1_id
mol0 = KBASE.ask(id0)
mol1 = KBASE.ask(id1)
atom_match0 = [int(i) for i in m.mcs_atom0.split(',')]
atom_match1 = [int(i) for i in m.mcs_atom1.split(',')]
ret.append(
self.deposit_to_kbase(id0, id1, atom_match0, atom_match1))
return ret
mol1 = KBASE.ask(id1)
mcs = KBASE.ask(id0).extract(atom_match0)
for i, e in enumerate(atom_match1, start=1):
mcs.atom_prop[i]["mapped_index"] = e
mcs.set_title(title)
mcs.set_id(mcs_id)
return mcs
if ("__main__" == __name__):
filenames = [
"xfer3.11.mol2",
"xfer3.12.mol2",
]
id_list = struc.read_n_files(filenames)
mol0 = KBASE.ask(id_list[0])
mol1 = KBASE.ask(id_list[1])
mcs = SchrodMcs(3)
mcs_id = mcs.search(mol0, mol1)[0]
mol_id = KBASE.ask(mcs_id, "mcs-parents")[0]
mcs_struc = KBASE.ask(mcs_id)[0]
mol_struc = KBASE.ask(mol_id)
out_fname = "out.mae"
if (os.path.isfile(out_fname)):
os.remove(out_fname)
mol_struc.write(out_fname)
mcs_struc.write(out_fname)
num_heavy_atoms = len( mcs0.heavy_atoms() )
num_light_atoms = len( mcs0.atom ) - num_heavy_atoms
KBASE.deposit_extra( mcs_id, "num_heavy_atoms", num_heavy_atoms )
KBASE.deposit_extra( mcs_id, "num_light_atoms", num_light_atoms )
return similarity.exp_delta( 2 * (orig_num_heavy_atoms - num_heavy_atoms), 0 )
# Example of a complex rule: A combination of a few simple rules (in case, they are Mcs, MinimumNumberOfAtom, and Cutoff).
# cutoff_simi = Cutoff( 0.2, Mcs( MinimumNumberOfAtom( 4 ) ) )
if ("__main__" == __name__) :
import struc
from mcs import SchrodMcs
from kbase import KBASE
filenames = ["xfer3.11.mol2", "xfer3.12.mol2",]
id_list = struc.read_n_files( filenames )
mol0 = KBASE.ask( id_list[0] )
mol1 = KBASE.ask( id_list[1] )
mcs = SchrodMcs( 3 )
mcs_id = mcs.search( mol0, mol1 )[0]
mol_id = KBASE.ask( mcs_id, "mcs_parents" )
print MCS.similarity( mol_id[0], mol_id[1] )
def gen_graph(mcs_ids, basic_rule, slack_rule, simi_cutoff, max_csize,
num_c2c):
"""
Generates and returns a graph according to the requirements.
@type mcs_ids: C{list} of C{str}
@param mcs_ids: A list of ids of the maximum substructures in C{KBASE}
@type rule: C{rule.Rule}
@param rule: The rule to determine the similarity score between two structures
@type simi_cutoff: C{float}
@param simi_cutoff: Cutoff of similarity scores. Values less than the cutoff are considered as 0.
@type max_csize: C{int}
@param max_csize: Maximum cluster size
@type num_c2c: C{int}
@param num_c2c: Number of cluster-to-cluster edges
"""
basic_graph = networkx.Graph()
all_ids = set()
fh = open("simiscore", "w") if (logging.getLogger().getEffectiveLevel()
== logging.DEBUG) else None
logging.info(" Calculating similarity scores...")
for id in mcs_ids:
#get the id for mol0 and mol1
id0, id1 = mcs.get_parent_ids(id)
#calculate the similarity scores for molecule pair
simi = basic_rule.similarity(id0, id1, mcs_id=id)
slack_simi = slack_rule.similarity(id0, id1, mcs_id=id)
KBASE.deposit_extra(id, "similarity", simi)
KBASE.deposit_extra(id, "slack_similarity", slack_simi)
all_ids.add(id0)
all_ids.add(id1)
if (fh):
print >> fh, simi
logging.info(" Calculating similarity scores... Done")
basic_graph.add_nodes_from(all_ids)
#create a complete graph
complete = create(basic_graph, mcs_ids, rule.Cutoff(0))
#delete connections with scores lower than simi_cutoff
desired = cutoff_graph(complete, simi_cutoff)
#get molecule clusters
clusters = sorted(networkx.connected_components(desired),
cmp=lambda x, y: len(y) - len(x))
logging.info(" Original number of clusters: %d" % len(clusters))
#break down big clusters
num_big_clusters = 0
for i, c in enumerate(clusters):
logging.info(" size of cluster #%02d: %d" % (i, len(c)), )
num_big_clusters += (len(c) > max_csize)
if (num_big_clusters and False):
logging.info(
" %d cluster(s) are too big. Break them into smaller ones. Reclustering..."
% num_big_clusters)
new_clusters = []
for c in clusters:
if (max_csize < len(c)):
new_clusters += break_cluster(desired.subgraph(c), simi_cutoff,
max_csize)
else:
new_clusters.append(c)
clusters = new_clusters
logging.info(" Reclustering... Done")
clusters = sorted(clusters, cmp=lambda x, y: len(y) - len(x))
n = len(clusters)
logging.info(" %d clusters in total" % n)
for i, c in enumerate(clusters):
logging.info(" size of cluster #%02d: %d" % (
i,
len(c),
))
# Optimizes the subgraphs.
logging.info(" Optimizing the subgraph of each cluster...")
new_desired = networkx.Graph()
for e in clusters:
sg = optimize_graph(complete.subgraph(e), desired.subgraph(e), "trim",
simi_cutoff)
new_desired = networkx.union(new_desired, sg)
desired = new_desired
logging.info(" Optimizing the subgraph of each cluster... Done")
# Connects the clusters.
unconnected_clusters = set(range(n))
while (unconnected_clusters and n > 1):
c2c_edges = []
cluster_index = unconnected_clusters.pop()
this_cluster = clusters[cluster_index]
other_clusters = copy.copy(clusters)
other_clusters.remove(this_cluster)
for e in other_clusters:
c2c_edges.extend(networkx.edge_boundary(complete, this_cluster, e))
if (len(c2c_edges) == 0):
logging.warn("WARNING: Cannot connect cluster #%d with others." %
(cluster_index, ))
logging.warn(
" If there should be connections, consider to adjust the rules to"
)
logging.warn(
" reduce 0-similarity assignments or loosen the MCS conditions."
)
continue
c2c_edges.sort(lambda x, y: cmp_edge(complete, x, y))
connected_clusters = set()
for k in range(-1, -num_c2c - 1, -1):
edge = c2c_edges[k]
node0 = edge[0]
node1 = edge[1]
simi = complete[node0][node1]["similarity"]
mcs_id = complete[node0][node1]["mcs_id"]
desired.add_edge(node0,
node1,
similarity=simi,
boundary=True,
mcs_id=mcs_id)
logging.warn(" boundary similarity = %f between '%s' and '%s'" % (
simi,
KBASE.ask(node0),
KBASE.ask(node1),
))
for e in unconnected_clusters:
if (node0 in clusters[e] or node1 in clusters[e]):
connected_clusters.add(e)
unconnected_clusters -= connected_clusters
return desired, clusters
def _similarity( self, id0, id1, **kwarg ) :
# Uses the first common substructure.
mcs_id = kwarg["mcs_id"]
mcs0 = mcs.get_struc( mcs_id ).copy()
mol0 = KBASE.ask( id0 )
mol1 = KBASE.ask( id1 )
orig_num_heavy_atoms = len( mcs0.heavy_atoms() )
# Deletes chiral atoms.
chiral_atoms = mcs0.chiral_atoms()
ring_atoms = mcs0. ring_atoms()
chiral_atoms.sort( reverse = True )
for atom_index in chiral_atoms :
if (atom_index in ring_atoms) :
#if the chiral atom is in a ring, delete atoms attached to it but not in ring.
bonded_atoms = set( mcs0.bonded_atoms( atom_index ) ) - ring_atoms
if (bonded_atoms) :
i = 0
n = 0
for atom in bonded_atoms :
cp = mcs0.copy()
cp.delete_atom( atom )
m = len( cp.atom )
if (m > n) :
i = atom
n = m
mcs0.delete_atom( i )
mcs0 = mcs0.copy()
else :
logging.warn( "WARNING: Cannot delete chiral atom #%d in structure: %s" % (atom_index, mcs0.title(),) )
else :
# If the chiral atom is not a ring atom, we simply delete it.
mcs0.delete_atom( atom_index )
mcs0 = mcs0.copy()
# If the deletion results in multiple unconnected fragments, we keep only the biggest one.
mcs0 = mcs0.copy()
atoms_to_delete = []
for e in mcs0.molecules()[1:] :
atoms_to_delete.extend( e )
mcs0.delete_atom( atoms_to_delete )
mcs0 = mcs0.copy()
partial_ring = self._delete_broken_ring( mol0, mol1, mcs0 )
mcs0 = mcs0.copy()
atoms_to_delete_2 = []
for e in mcs0.molecules()[1:] :
atoms_to_delete_2.extend( e )
mcs0.delete_atom( atoms_to_delete_2 )
#Here is different from schrodinger's method. Since openeye cannot save mcs searching results as smart strings. Using smiles stirng instead for later layout.
smiles0 = mcs0.smiles()
#Arbitrarily set the simles0 = smiles1 (do not considering the mcs searching difference between mol0 matching to mol1 vs mol1 matching to mol0)
smiles1 = smiles0
KBASE.deposit_extra( mcs_id, "trimmed-mcs", {id0:smiles0, id1:smiles1,} )
KBASE.deposit_extra( mcs_id, "partial_ring", len( partial_ring ) )
KBASE.deposit_extra( mcs_id, "layout_mcs", smiles0 )
num_heavy_atoms = len( mcs0.heavy_atoms() )
num_light_atoms = len( mcs0.atom ) - num_heavy_atoms
KBASE.deposit_extra( mcs_id, "num_heavy_atoms", num_heavy_atoms )
KBASE.deposit_extra( mcs_id, "num_light_atoms", num_light_atoms )
return similarity.exp_delta( 2 * (orig_num_heavy_atoms - num_heavy_atoms), 0 )
def _similarity( self, id0, id1, **kwarg ) :
# Uses the first common substructure.
mcs_id = kwarg["mcs_id"]
mcs0 = mcs.get_struc( mcs_id ).copy()
mol0 = KBASE.ask( id0 )
mol1 = KBASE.ask( id1 )
orig_num_heavy_atoms = len( mcs0.heavy_atoms() )
partial_ring = self._delete_broken_ring( mol0, mol1, mcs0 )
# Deletes chiral atoms.
chiral_atoms = mcs0.chiral_atoms()
ring_atoms = mcs0. ring_atoms()
chiral_atoms.sort( reverse = True )
for atom_index in chiral_atoms :
if (atom_index in ring_atoms) :
bonded_atoms = set( mcs0.bonded_atoms( atom_index ) ) - ring_atoms
if (bonded_atoms) :
i = 0
n = 0
for atom in bonded_atoms :
cp = mcs0.copy()
cp.delete_atom( atom )
m = len( cp.atom )
if (m > n) :
i = atom
n = m
mcs0.delete_atom( i )
else :
# If the chiral atom is not a ring atom, we simply delete it.
mcs0.delete_atom( atom_index )
# If the deletion results in multiple unconnected fragments, we keep only the biggest one.
atoms_to_delete = []
for e in mcs0.molecules()[1:] :
atoms_to_delete.extend( e )
mcs0.delete_atom( atoms_to_delete )
# Gets the SMARTS for the trimmed structure.
atom_list0 = []
atom_list1 = []
for e in mcs0.atom_prop[1:] :
atom_list0.append( e[ "orig_index"] )
atom_list1.append( e["mapped_index"] )
smarts0 = mol0.smarts( atom_list0 )
try :
smarts1 = mol1.smarts( atom_list1 )
except ValueError :
smarts1 = ""
KBASE.deposit_extra( mcs_id, "trimmed-mcs", {id0:smarts0, id1:smarts1,} )
KBASE.deposit_extra( mcs_id, "partial_ring", len( partial_ring ) )
KBASE.deposit_extra( mcs_id, "layout_mcs", mcs0.smiles() )
num_heavy_atoms = len( mcs0.heavy_atoms() )
num_light_atoms = len( mcs0.atom ) - num_heavy_atoms
KBASE.deposit_extra( mcs_id, "num_heavy_atoms", num_heavy_atoms )
KBASE.deposit_extra( mcs_id, "num_light_atoms", num_light_atoms )
return similarity.exp_delta( 2 * (orig_num_heavy_atoms - num_heavy_atoms), 0 )