class Pair:
def __init__(self, receptordb, liganddb):
self.ligand = Ligand(liganddb)
self.receptor = Receptor(receptordb)
def foreachPair(self, pid, sink):
self.ligand.foreachId(lambda cid: sink(pid, cid))
def foreach(self, sink, criteria=''):
self.receptor.foreachId(lambda pid: self.foreachPair(pid, sink))
def size(self, rcriteria='1=1', lcriteria='1=1'):
ligands = self.ligand.size(lcriteria)
receptors = self.receptor.size(rcriteria)
return ligands * receptors
def close(self):
self.ligand.close()
self.receptor.close()
def main():
# Parse command-line args.
args = parse_args()
# Make the libraries used by the Ligand class shut up.
Ligand(None).quiet()
# Set random seed.
numpy.random.seed(args.seed)
# Perform the GA search.
gnn = init_gnn(args)
# Perform the GA search.
ga_best = ga_search(args, gnn)
def run():
ligand = None
protein = request.form['protein']
data = dict(request.form)
if request.files:
file_obj = request.files
for f in file_obj:
ligand = request.files.get(f)
word = Ligand.save(ligand, app.instance_path)
return render_template("results.html", ligand=word, protein=protein)
def init_gnn(args):
# Setup gnn fitness oracle.
protein = load_protein_file(args.protein)
ligand = Ligand("CCC").get_dict()
target = numpy.array([0.0, 1.0], dtype=numpy.float32)
target = {
'n_node': 0,
'n_edge': 0,
'nodes': None,
'edges': None,
'senders': None,
'receivers': None,
'globals': target
}
ops = build_gnn(args, (protein, ligand, target))
config = tf.ConfigProto(allow_soft_placement=True,
log_device_placement=False)
config.gpu_options.allow_growth = False
sess = tf.Session(config=config)
sess.run(tf.global_variables_initializer())
saver = tf.train.Saver()
saver.restore(sess, args.model)
return (sess, ops)
def postLigand(ligand):
word = Ligand.save(ligand, app.instance_path)
return word
def pdf():
ligand = {"EGFR": 1, "EGFE": 2}
ml = {"Support Vector Machines": [0.5, 0.5], "KNN": [0.4, 0.6]}
pdf = Ligand.makePDF(ligand, ml)
return app.send_static_file(pdf)
def Loadpdb(pdb=None, hetatm= True, verbose=False):
try:
assert(pdb != None) #Check if filehandle to PDB file is passed
except AssertionError:
sys.exit("**No filehandle passed**. Pass a filehandle (to a pdb file) as an argument to Loadpdb.")
AtomNumber=0 #Keeps track of atom indices (assigned in the order atoms listed in input file)
mol_data={} #Key: molid; Value: Molecule_Type Object; Keep track of different molecules (different chains or molecule type) in input structure
first_res =True #To identify molecule type of every molecule in input structure and accordingly define Molecule object.
Prev_res=0 # to keep track of residue change
Prev_chain='aa' # to keep track of chain change in HETATM section
atmTohet = True #To determine transition from ATOM to HETATM record
frame_tag = '' # To keep track of multi-frame entry (multiple entry for same molecule type with same chain id)
'''Load the PDB structure file'''
for line in pdb:
if line[0:4]=="ATOM" and line[12:16].upper().strip() not in ["OXT"]:
AtomNumber+=1
AtomName, ResName, Chain, ResNo, CordX, CordY, CordZ, Occ, Bfac = Pdbcordsec(line)
atm = Atom(AtomName, AtomNumber, ResName, Chain, ResNo, CordX, CordY, CordZ)
#Check for unrecognized residue and new molecule
if not first_res:
if ResName.lower() not in Mol_types['protein'] + Mol_types['lipid'] + Mol_types['ligand']:
print "*** Unrecognized residue name: "+ ResName+ " ***.\nAdded %s as Ligand." % ResName
sys.exit("In file configstruc.py: Add missing residue name("+ ResName+ ") to appropriate molecule in Mol_types")
if ResNo != Prev_res or (Prev_chain != Chain and mol.molecule_type().lower()=='ligand'):
if mol.molecule_type().lower()=='ligand':
mol_data[Molecule.molid] = deepcopy(mol) #copy mol object into dictionary
first_res = True
elif Prev_chain != Chain or ResName.lower() not in Mol_types[mol.molecule_type().lower()]: #Either Chain is different or New residue doesn't belong to current molecule type
mol_data[Molecule.molid] = deepcopy(mol) #copy mol object into dictionary
first_res = True
elif frame_tag.lower() in ['endmdl', 'ter', 'end'] and Prev_chain == Chain: #Different molecule (of same molecule type) with same chain id; as in trajectory frames
mol_data[Molecule.molid] = deepcopy(mol) #copy mol object into dictionary
first_res = True
if first_res: #Initialize mol for new chain or molecule
if ResName.lower() in Mol_types['protein']:
mol = Protein()
elif ResName.lower() in Mol_types['lipid']:
mol = Lipid()
elif ResName.lower() in Mol_types['ligand']:
mol = Ligand()
else:
print "*** Unrecognized residue name: "+ ResName+ " ***.\n Cannot initialize Molecule object."
sys.exit("In file configstruc.py: Add missing residue name("+ ResName+ ") to appropriate molecule in Mol_types")
first_res = False
frame_tag = ''
mol.AddToResidue(atom=atm, occ=Occ, bfac=Bfac)
mol.atmidx.append(AtomNumber)
Prev_res= ResNo
Prev_chain=Chain
elif line[0:6]=="HETATM" and hetatm == True:
if atmTohet:
mol_data[Molecule.molid] = deepcopy(mol) #copy mol object into dictionary
first_res = True
atmTohet = False
AtomNumber+=1
AtomName, ResName, Chain, ResNo, CordX, CordY, CordZ, Occ, Bfac = Pdbcordsec(line)
atm = Atom(AtomName, AtomNumber, ResName, Chain, ResNo, CordX, CordY, CordZ)
#Check for new ligand molecule
if not first_res and (ResNo != Prev_res or Prev_chain != Chain):
mol_data[Molecule.molid] = deepcopy(mol) #copy mol object into dictionary
first_res = True
#Initialize mol for new chain or molecule
if first_res:
if ResName.lower() in Mol_types['ligand']:
mol = Ligand()
else:
print "*** Unrecognized residue name: "+ ResName+ " ***.\n Cannot initialize Ligand object."
sys.exit("In file configstruc.py: Add missing residue name ("+ ResName+ ") to ligand molecule in Mol_types")
first_res = False
mol.AddToResidue(atom=atm, occ=Occ, bfac=Bfac)
mol.atmidx.append(AtomNumber)
Prev_res= ResNo
Prev_chain=Chain
elif line[0:3].lower() in ["ter", "end"] or line[0:6].lower() == "endmdl":
frame_tag = line[0:3]
#append the last mol object to mol_data
mol_data[Molecule.molid] = deepcopy(mol) #copy mol object into dictionary
if verbose:
print "Number of molecules in input file: ", len(mol_data), "\n"
#Update mol_data[molid].nor, mol_data[molid].resids, and check for chain breaks in non-ligand molecules
for key in sorted(mol_data):
if verbose:
print "Molid:", key,"Molecule_Type:",mol_data[key].molecule_type()
mol_data[key].resids = sorted(mol_data[key].residue)
mol_data[key].nor = len(mol_data[key].resids)
if mol_data[key].molecule_type().lower() != 'ligand':
#Check for chain breaks in non-ligand molecule
resids_diff=numpy.array(mol_data[key].resids[1:]) - numpy.array(mol_data[key].resids[:-1])
if mol_data[key].nor != (numpy.sum(resids_diff)+1):
break_indices = numpy.where(resids_diff > 1)
print "Chain break encountered in molecule",key, "at residue positions: "
for res in break_indices[0]:
print mol_data[key].resids[res],
print "\n"
mol_data[key].chain_break = True
return mol_data
import os
from parameter import Parameter
from ligand import Ligand
from receptor import Receptor
from path import Path
def print_pairs(pid, ligand):
ligand.foreachId(lambda cid: print(pid, cid, sep='\t'))
p = Parameter()
pt = Path(p)
ligand = Ligand(pt.data + p._('project.wizard.ligand.db'))
receptor = Receptor(pt.data + p._('project.wizard.receptor.db'))
receptor.foreachId(lambda id: print_pairs(id, ligand))
ligand.close()
receptor.close()
import os
from pdbqtdaemon import PDBQTDaemon
from parameter import Parameter
from ligand import Ligand
from path import Path
process_name = 'prepare_ligand'
p = Parameter()
pt = Path(p)
user = p._('user')
payload = p.i('daemon.sbatch.payload')
db = Ligand(pt.liganddb)
class PrepareLigand(PDBQTDaemon):
def __init__(self):
PDBQTDaemon.__init__(self, user, db, pt, payload, process_name)
def molecule_done(self, id):
return os.path.isfile(pt.ligandpdbqt(id))
def prepare(self, id, templ):
os.chdir(self.path.docking_ligand)
log_name = '{}/{}_{}.out'.format(self.path.log, self.proc_name, id)
templ = self.template.format(log_name, self.path.ligand(id), id)
target = '{}_{}_{}.sbatch'.format(self.path.project_name,
self.proc_name, id)
open(target, 'w').write(templ)
from parameter import Parameter
from ligand import Ligand
from receptor import Receptor
def line(id, ligand):
ids = [str(id)]
ligand.foreachId(lambda id: ids.append(str(id)))
return '\t'.join(ids)
p = Parameter()
ligand = Ligand(p._('project.ligand.db'))
receptor = Receptor(p._('project.receptor.db'))
receptor.foreachId(lambda id: print(line(id, ligand)))
ligand.close()
receptor.close()
def Loadpdb(pdb=None, hetatm=True, verbose=False):
try:
assert (pdb != None) #Check if filehandle to PDB file is passed
except AssertionError:
sys.exit(
"**No filehandle passed**. Pass a filehandle (to a pdb file) as an argument to Loadpdb. "
)
AtomNumber = 0 #Keeps track of atom indices (assigned in the order atoms listed i input file)
mol_data = {
} #Key: molid; Keep track of different molecules (different chains or molecule type) in input structure
check_het = False #To keep track of new Hetero residue
first_res = True #To identify molecule type of every molecule in input structure and accordingly define Molecule object.
Prev_res = 0 # to keep track of residue change in HETATM section; a new Molecule object is assigned for every residue.
Prev_chain = 'a' # to keep track of chain change in HETATM section
'''Load the PDB structure file'''
for line in pdb:
if line[0:4] == "ATOM" and line[12:16].upper().strip() not in ["OXT"]:
AtomNumber += 1
AtomName, ResName, Chain, ResNo, CordX, CordY, CordZ, Occ, Bfac = Pdbcordsec(
line)
atm = Atom(AtomName, AtomNumber, ResName, Chain, ResNo, CordX,
CordY, CordZ)
if first_res: #Initialize mol for new chain or molecule
if ResName.lower() in Mol_types['protein']:
mol = Protein()
elif ResName.lower() in Mol_types['ligand']:
mol = Ligand()
else:
print "*** Unrecognized residue name: " + ResName + " ***.\n Cannot initialize Molecule object."
sys.exit(
"In file configstruc.py: Add missing residue name(" +
ResName + ") to appropriate molecule in Mol_types")
first_res = False
mol.AddToResidue(atom=atm, occ=Occ, bfac=Bfac)
mol.atmidx.append(AtomNumber)
elif line[0:3] == "TER":
mol_data[Molecule.molid] = deepcopy(
mol) #copy mol object into dictionary
first_res = True # mol object will be initialized to molecule type of next molecule
elif line[0:6] == "HETATM" and hetatm == True:
AtomNumber += 1
AtomName, ResName, Chain, ResNo, CordX, CordY, CordZ, Occ, Bfac = Pdbcordsec(
line)
atm = Atom(AtomName, AtomNumber, ResName, Chain, ResNo, CordX,
CordY, CordZ)
#Check for new molecule
if (
ResNo != Prev_res or Prev_chain != Chain
) and check_het == True: #For first HETATM check_het is always False
mol_data[Molecule.molid] = deepcopy(
mol) #copy mol object into dictionary
first_res = True
#Initialize mol for new chain or molecule
if first_res:
if ResName.lower() in Mol_types['ligand']:
mol = Ligand()
else:
print "*** Unrecognized residue name: " + ResName + " ***.\n Cannot initialize Ligand object."
sys.exit(
"In file configstruc.py: Add missing residue name (" +
ResName + ") to ligand molecule in Mol_types")
first_res = False
mol.AddToResidue(atom=atm, occ=Occ, bfac=Bfac)
mol.atmidx.append(AtomNumber)
if Prev_res == 0:
check_het = True
Prev_res = ResNo
Prev_chain = Chain
if hetatm: #If HETATM record was added; append the last hetero residue object to mol_data
mol_data[Molecule.molid] = deepcopy(
mol) #copy mol object into dictionary
if verbose:
print "Number of molecules in input file: ", len(mol_data), "\n"
#Update mol_data[molid].nor, mol_data[molid].resids, and check for chain breaks in non-ligand molecules
for key in sorted(mol_data):
if verbose:
print "Molid:", key, "Molecule_Type:", mol_data[key].molecule_type(
)
if mol_data[key].molecule_type().lower() != 'ligand':
mol_data[key].resids = sorted(mol_data[key].residue)
mol_data[key].nor = len(mol_data[key].resids)
#Check for chain breaks in protein
resids_diff = numpy.array(mol_data[key].resids[1:]) - numpy.array(
mol_data[key].resids[:-1])
if mol_data[key].nor != (numpy.sum(resids_diff) + 1):
break_indices = numpy.where(resids_diff > 1)
print "Chain break encountered in molecule", key, "at residue positions: "
for res in break_indices[0]:
print mol_data[key].resids[res],
print "\n"
return mol_data
import os
from parameter import Parameter
from ligand import Ligand
p = Parameter()
dir = p._('project.dir') + '/'
project_dir = dir + p._('project.wizard.create.name') + '/'
data_dir = project_dir + 'data/'
molecule = Ligand(data_dir + p._('project.wizard.ligand.db'))
columns = p._('project.wizard.ligand_report.columns')
molecule.foreachRecord(columns, lambda id, fields: print(id, fields, sep='\t'))
molecule.close()
import os
from parameter import Parameter
from ligand import Ligand
p = Parameter()
#locate tree
dir = p._('project.dir')
name = p._('project.wizard.create.name')
project_dir = dir + '/' + name
data_dir = project_dir + '/data'
filename = '{}/stage/{}'.format(project_dir, p._('project.wizard.load.ligand'))
#import data
molecule = Ligand(data_dir + '/' + p._('project.wizard.ligand.db'))
molecule.clear()
for id in open(filename):
id = id.strip('\n')
molecule.add(id)
molecule.commit()
# download ligand structures
def save(id, s, format, dir):
filename = '{}/{}.{}'.format(dir, id, format)
with open(filename, 'w') as file:
file.write(s)
print(id)
def __init__(self, receptordb, liganddb):
self.ligand = Ligand(liganddb)
self.receptor = Receptor(receptordb)
def getSampleScores(protein, active):
dict = Ligand.getSample(protein, active)
return make_response(jsonify(dict))
def ga_search(args, gnn):
# Create initial population.
protein = load_protein_file(args.protein)
population = [Ligand(smiles) for smiles in args.initpop.split(',')]
population += [
population[0].clone() for i in range(args.popsz - len(population))
]
for ind in population[1:]:
ind.mutate(args.mu)
# Generation loop.
print("Start genetic algorithm search (%d):" % args.gens)
best = population[0]
best_fit = WORST_ERROR
images = []
for gen in range(args.gens):
# Stdout print.
print(" Generation %d:" % gen)
print(" PopSize: %d" % len(population))
# Compute fitness.
tstart = time.time()
fits = eval_inds(population, gnn, protein)
tfend = time.time()
viable = sum(fit != WORST_ERROR for fit in fits)
# Viability selection.
if viable is 0:
print(" Population went extinct!")
break
print(" nViable: %d" % viable)
indexes = list(range(len(fits)))
indexes.sort(key=fits.__getitem__)
fits = list(map(fits.__getitem__, indexes))
population = list(map(population.__getitem__, indexes))
population = population[0:int(args.popsz * args.sigma)]
elite, efits = ([], [])
for ndx, ind in enumerate(population):
if ind.smiles not in elite:
elite.append(ind.smiles)
efits.append(fits[ndx])
if len(elite) == 3:
break
population = [
ind for i, ind in enumerate(population) if fits[i] != WORST_ERROR
]
fits = fits[0:len(population)]
# Some initial stats.
if fits[0] < best_fit:
best_fit = fits[0]
best = population[0]
if args.img:
images.append("%s/best_%d.png" % (args.out, gen))
best.write_image(images[-1])
write_gif(images, "%s/best.gif" % (args.out))
print(" Fitness: %f" % (sum(fits) / len(fits)))
print(" BestFit: %f" % (best_fit))
print(" Best: %s" % (best.smiles))
print(" Elite:")
for ndx, ind in enumerate(elite):
print(" (%.4f) %s" % (efits[ndx], ind))
# Reproduction.
opop = len(population)
population.append(population[0].clone())
while len(population) < args.popsz:
# Crossover.
indi = population[numpy.random.randint(0, opop)]
indj = population[numpy.random.randint(0, opop)]
inds = indi.mate(indj)
# Mutation.
for ind in inds:
ind.mutate(args.mu)
population.append(ind)
# End of generation timing and print.
tend = time.time()
print(" Time: %.2f s (fit %.3f, ga %.3f)" %
(tend - tstart, tfend - tstart, tend - tfend))
# Return best found.
return best
from path import Path
from ligand import Ligand
p = Parameter()
pt = Path(p)
#locate tree
dir = p._('project.dir')
name = p._('project.wizard.create.name')
project_dir = dir + '/' + name
data_dir = project_dir + '/data'
filename = '{}/stage/{}'.format(project_dir, p._('project.wizard.load.ligand'))
#import data
molecule = Ligand(pt.liganddb)
molecule.clear()
for id in open(filename):
id = id.strip('\n')
molecule.add(id)
molecule.commit()
# download ligand structures
def instage(id, pt):
return pt.filename('{}.pdb'.format(id), pt.stage, '*.pdb') != None
def copy(id, dir, pt, pids):