def parse_index(path, index):
'''
'''
regexp = r"""^
(?P<pdb>\w{4})\s+
(?P<resolution>\d[.]\d{2}|NMR)\s+
(?P<year>\d{4})\s+
(?P<pKx>\d{1,2}[.]\d{2})\s+
(?P<type>\w{2,4})
(?P<relation>[<>=~]{1,2})
(?P<value>\d+[.]\d+|\d+)
(?P<unit>\w{2}).+"""
pattern = re.compile(regexp, re.VERBOSE)
data = {}
for line in open(index):
if not line.startswith('#'):
match = pattern.match(line)
# PRINT A WARNING IF REGULAR EXPRESSION FAILED ON A LINE
if not match:
logger.warn("Could not parse line: {0}".format(line))
continue
rowdata = match.groupdict()
pdb = rowdata.pop('pdb')
data[pdb] = rowdata
return data
def parse_index(path, index):
'''
'''
regexp = r"""^
(?P<pdb>\w{4})\s+
(?P<resolution>\d[.]\d{2}|NMR)\s+
(?P<year>\d{4})\s+
(?P<pKx>\d{1,2}[.]\d{2})\s+
(?P<type>\w{2,4})
(?P<relation>[<>=~]{1,2})
(?P<value>\d+[.]\d+|\d+)
(?P<unit>\w{2}).+"""
pattern = re.compile(regexp, re.VERBOSE)
data = {}
for line in open(index):
if not line.startswith('#'):
match = pattern.match(line)
# PRINT A WARNING IF REGULAR EXPRESSION FAILED ON A LINE
if not match:
logger.warn("Could not parse line: {0}".format(line))
continue
rowdata = match.groupdict()
pdb = rowdata.pop('pdb')
data[pdb] = rowdata
return data
def parse_index(path, index):
'''
parse files:
INDEX_general_PL_data.2015
INDEX_refined_data.2015
INDEX_core_data.2015
for example:
{'1mhw':{'year': '2002',
'resolution': '1.90',
'type': 'Ki',
'value': '45',
'relation': '=',
'pKx': '7.35',
'unit': 'nM'}
}
'''
regexp = r"""^
(?P<pdb>\w{4})\s+
(?P<resolution>\d[.]\d{2}|NMR)\s+
(?P<year>\d{4})\s+
(?P<pKx>\d{1,2}[.]\d{2})\s+
(?P<type>\w{2,4})
(?P<relation>[<>=~]{1,2})
(?P<value>\d+[.]\d+|\d+)
(?P<unit>\w{2}).+"""
pattern = re.compile(regexp, re.VERBOSE)
data = {}
for line in open(index):
if not line.startswith('#'):
match = pattern.match(line)
# PRINT A WARNING IF REGULAR EXPRESSION FAILED ON A LINE
if not match:
logger.warn("Could not parse line: {0}".format(line))
continue
rowdata = match.groupdict()
pdb = rowdata.pop('pdb')
data[pdb] = rowdata
return data
def sift_descriptor(protein, ligand, binsize=0.0):
"""
Calculates a descriptor of the protein-ligand complex as the sum of the structural
interaction fingerprints (SIFTs) of all interacting atoms.
Parameters
----------
protein_path: str
Path to the PDB structure of the protein.
ligand_path: str
Path to the structure of the ligand, must be readable by Open Babel.
binsize: float
Size of the distance bins in Angstrom that will be used to bin the contacts.
The total range will be from 1.0 to <cutoff> + <binsize> in <binsize> steps.
Returns
-------
descriptor: numpy.ndarray
The shape of the descriptor array will be 1D equal to the number of contact
types or 2D (number of bins, number of contact types) if a binsize was given.
"""
# SUPPRESS OPENBABEL WARNINGS
pybel.ob.obErrorLog.StopLogging()
# ELEMENT TABLE TO DETERMINE VDW AND COVALENT BONDS
et = OBElementTable()
# CREDO DESCRIPTOR LABELS
interaction_types = ['covalent','vdw_clash','vdw','proximal','hbond','weak_hbond',
'xbond','ionic','metal_complex','aromatic','hydrophobic',
'carbonyl']
numcols = len(interaction_types)
# GENERATE THE DISTANCE BINS
if binsize:
# get the distance bins for the given cutoff and bin size
bins = get_distance_bins(config['cutoff'], binsize)
offset = bins.size + 1
# DEBUG DISTANCE BINS
logger.debug("The distance bins in Angstrom are {0}.".format(bins))
# NUMBER OF TOTAL COLUMNS IN DESCRIPTOR
numcols *= (bins.size + 1)
labels = []
# CREATE A COLUMN FOR EACH ELEMENT PAIR AND DISTANCE BIN
for interaction_type in interaction_types:
for i in range(len(bins) + 1):
label = "{0}-B{1}".format(interaction_type, i)
labels.append(label)
# LABEL WITHOUT BINS
else: labels = interaction_types
# DESCRIPTOR THAT WILL CONTAIN THE SUM OF ALL ELEMENT-ELEMENT INTERACTIONS
descriptor = numpy.zeros(numcols, dtype=int)
# GET THE ATOM TYPES FOR THE LIGAND
# CALCULATED ON THE FLY
lig_atom_types = ob.get_atom_types(ligand, config)
contacts = get_contacts(protein, ligand, config['cutoff'])
# ITERATE THROUGH CONTACT PAIRS AND DETERMINE SIFT
for hetatm, hetatm_contacts in contacts:
# GET THE ATOM TYPES FOR THE HETATM
hetatm_types = lig_atom_types[hetatm.GetIdx()]
# GET ATOM RADII FOR THE LIGAND ATOM
hetatm_cov = et.GetCovalentRad(hetatm.GetAtomicNum())
hetatm_vdw = et.GetVdwRad(hetatm.GetAtomicNum())
# ITERATE THROUGH ALL THE CONTACTS THE HETATM HAS
for atom, distance in hetatm_contacts:
# INITIALIZE STRUCTURAL INTERACTION FINGERPRINT
sift = numpy.zeros(descriptor.size)
residue = atom.GetResidue()
res_name = residue.GetName()[:3]
# IGNORE WATER RESIDUES
if res_name == 'HOH': continue
# GET ATOM TYPES FOR THE PROTEIN ATOM
try:
atom_types = res_atom_types[res_name][residue.GetAtomID(atom).strip()]
except KeyError:
logger.warn("Cannot find atom types for {} {}."
.format(res_name, residue.GetAtomID(atom).strip()))
continue
sum_cov = hetatm_cov + et.GetCovalentRad(atom.GetAtomicNum())
sum_vdw = hetatm_vdw + et.GetVdwRad(atom.GetAtomicNum())
# BIN INTERACTIONS
if binsize:
# GET THE BIN THIS CONTACT BELONGS IN
# DIGITIZE TAKES AN ARRAY-LIKE AS INPUT
bin_id = numpy.digitize([distance,], bins)[0] + 1
else:
offset = 1
bin_id = 0
# COVALENT BOND - SHOULD NOT OCCUR IN PDBBIND
if distance <= sum_cov: sift[0 * offset + bin_id] = 1
# VAN DER WAALS CLASH
elif distance <= sum_vdw: sift[1 * offset + bin_id] = 1
# VAN DER WAALS CONTACT
elif distance <= sum_vdw + 0.5: sift[2 * offset + bin_id] = 1
# PROXIMAL
else: sift[3 * offset + bin_id] = 1
if interactions.is_hbond(hetatm,hetatm_types,atom,atom_types,distance): sift[4 * offset + bin_id] = 1
if interactions.is_weak_hbond(hetatm,hetatm_types,atom,atom_types,distance): sift[5 * offset + bin_id] = 1
if interactions.is_xbond(hetatm,hetatm_types,atom,atom_types,distance): sift[6 * offset + bin_id] = 1
if interactions.is_ionic(hetatm,hetatm_types,atom,atom_types,distance): sift[7 * offset + bin_id] = 1
if interactions.is_metal_complex(hetatm,hetatm_types,atom,atom_types,distance): sift[8 * offset + bin_id] = 1
if interactions.is_aromatic(hetatm,hetatm_types,atom,atom_types,distance): sift[9 * offset + bin_id] = 1
if interactions.is_hydrophobic(hetatm,hetatm_types,atom,atom_types,distance): sift[10 * offset + bin_id] = 1
if interactions.is_carbonyl(hetatm,hetatm_types,atom,atom_types,distance): sift[11 * offset + bin_id] = 1
descriptor += sift
if binsize: sum_descriptor_bins(descriptor, bins)
return descriptor, labels
def sift_descriptor(protein, ligand, binsize=0.0):
"""
Calculates a descriptor of the protein-ligand complex as the sum of the structural
interaction fingerprints (SIFTs) of all interacting atoms.
Parameters
----------
protein_path: str
Path to the PDB structure of the protein.
ligand_path: str
Path to the structure of the ligand, must be readable by Open Babel.
binsize: float
Size of the distance bins in Angstrom that will be used to bin the contacts.
The total range will be from 1.0 to <cutoff> + <binsize> in <binsize> steps.
Returns
-------
descriptor: numpy.ndarray
The shape of the descriptor array will be 1D equal to the number of contact
types or 2D (number of bins, number of contact types) if a binsize was given.
"""
# SUPPRESS OPENBABEL WARNINGS
pybel.ob.obErrorLog.StopLogging()
# ELEMENT TABLE TO DETERMINE VDW AND COVALENT BONDS
et = OBElementTable()
# CREDO DESCRIPTOR LABELS
interaction_types = [
'covalent', 'vdw_clash', 'vdw', 'proximal', 'hbond', 'weak_hbond',
'xbond', 'ionic', 'metal_complex', 'aromatic', 'hydrophobic',
'carbonyl'
]
numcols = len(interaction_types)
# GENERATE THE DISTANCE BINS
if binsize:
# get the distance bins for the given cutoff and bin size
bins = get_distance_bins(config['cutoff'], binsize)
offset = bins.size + 1
# DEBUG DISTANCE BINS
logger.debug("The distance bins in Angstrom are {0}.".format(bins))
# NUMBER OF TOTAL COLUMNS IN DESCRIPTOR
numcols *= (bins.size + 1)
labels = []
# CREATE A COLUMN FOR EACH ELEMENT PAIR AND DISTANCE BIN
for interaction_type in interaction_types:
for i in range(len(bins) + 1):
label = "{0}-B{1}".format(interaction_type, i)
labels.append(label)
# LABEL WITHOUT BINS
else:
labels = interaction_types
# DESCRIPTOR THAT WILL CONTAIN THE SUM OF ALL ELEMENT-ELEMENT INTERACTIONS
descriptor = numpy.zeros(numcols, dtype=int)
# GET THE ATOM TYPES FOR THE LIGAND
# CALCULATED ON THE FLY
lig_atom_types = ob.get_atom_types(ligand, config)
contacts = get_contacts(protein, ligand, config['cutoff'])
# ITERATE THROUGH CONTACT PAIRS AND DETERMINE SIFT
for hetatm, hetatm_contacts in contacts:
# GET THE ATOM TYPES FOR THE HETATM
hetatm_types = lig_atom_types[hetatm.GetIdx()]
# GET ATOM RADII FOR THE LIGAND ATOM
hetatm_cov = et.GetCovalentRad(hetatm.GetAtomicNum())
hetatm_vdw = et.GetVdwRad(hetatm.GetAtomicNum())
# ITERATE THROUGH ALL THE CONTACTS THE HETATM HAS
for atom, distance in hetatm_contacts:
# INITIALIZE STRUCTURAL INTERACTION FINGERPRINT
sift = numpy.zeros(descriptor.size)
residue = atom.GetResidue()
res_name = residue.GetName()[:3]
# IGNORE WATER RESIDUES
if res_name == 'HOH': continue
# GET ATOM TYPES FOR THE PROTEIN ATOM
try:
atom_types = res_atom_types[res_name][residue.GetAtomID(
atom).strip()]
except KeyError:
logger.warn("Cannot find atom types for {} {}.".format(
res_name,
residue.GetAtomID(atom).strip()))
continue
sum_cov = hetatm_cov + et.GetCovalentRad(atom.GetAtomicNum())
sum_vdw = hetatm_vdw + et.GetVdwRad(atom.GetAtomicNum())
# BIN INTERACTIONS
if binsize:
# GET THE BIN THIS CONTACT BELONGS IN
# DIGITIZE TAKES AN ARRAY-LIKE AS INPUT
bin_id = numpy.digitize([
distance,
], bins)[0] + 1
else:
offset = 1
bin_id = 0
# COVALENT BOND - SHOULD NOT OCCUR IN PDBBIND
if distance <= sum_cov:
sift[0 * offset + bin_id] = 1
# VAN DER WAALS CLASH
elif distance <= sum_vdw:
sift[1 * offset + bin_id] = 1
# VAN DER WAALS CONTACT
elif distance <= sum_vdw + 0.5:
sift[2 * offset + bin_id] = 1
# PROXIMAL
else:
sift[3 * offset + bin_id] = 1
if interactions.is_hbond(hetatm, hetatm_types, atom, atom_types,
distance):
sift[4 * offset + bin_id] = 1
if interactions.is_weak_hbond(hetatm, hetatm_types, atom,
atom_types, distance):
sift[5 * offset + bin_id] = 1
if interactions.is_xbond(hetatm, hetatm_types, atom, atom_types,
distance):
sift[6 * offset + bin_id] = 1
if interactions.is_ionic(hetatm, hetatm_types, atom, atom_types,
distance):
sift[7 * offset + bin_id] = 1
if interactions.is_metal_complex(hetatm, hetatm_types, atom,
atom_types, distance):
sift[8 * offset + bin_id] = 1
if interactions.is_aromatic(hetatm, hetatm_types, atom, atom_types,
distance):
sift[9 * offset + bin_id] = 1
if interactions.is_hydrophobic(hetatm, hetatm_types, atom,
atom_types, distance):
sift[10 * offset + bin_id] = 1
if interactions.is_carbonyl(hetatm, hetatm_types, atom, atom_types,
distance):
sift[11 * offset + bin_id] = 1
descriptor += sift
if binsize: sum_descriptor_bins(descriptor, bins)
return descriptor, labels