def _calc_residue_dist(self, residue_one, residue_two, dist_atoms='CA'):
"""Returns the C-alpha distance between two residues"""
if not Polypeptide.is_aa(residue_one) or not Polypeptide.is_aa(residue_two):
return np.nan
dist_atom_1 = dist_atoms if dist_atoms in residue_one else 'CA'
dist_atom_2 = dist_atoms if dist_atoms in residue_two else 'CA'
try:
diff_vector = residue_one[dist_atom_1].coord - residue_two[dist_atom_2].coord
except KeyError:
return np.nan
return np.sqrt(np.sum(diff_vector * diff_vector))
def MutationsDict(file, positions=None):
"""Get dictionary with lists of mutations per position in protein, ignore
positions without residue in pdb file.
Parameters:
file (string): pdb file to get mutations from
positions: list of tuples of the form (chain, first, last) for positions
to mutate for all other aminoacids. If None, mutates all
positions in all chains
Returns:
dict with keys :aa:chain:position, each containing lists with
:aa:chain:position:mutated_aa for all mutations
"""
# Sorted list of one letter amino acids
AA = list(Bio.PDB.Polypeptide.aa1)
# Generate model of original pdb file
model = it.Pmolecule(file).model
# Dict to store mutations
mutations = dict()
if positions:
for chain_id, first, last in positions:
# Get chain corresponding to chain_id given
chain = next(chain for chain in model.get_chains()
if chain.id == chain_id)
for residue in chain:
if pp.is_aa(residue):
code = pp.three_to_one(residue.get_resname())
position = residue.id[1]
prefix = code + chain_id + str(position)
# Only save positions between first and last
if position in range(first, last + 1):
mutations[prefix] = [
prefix + aa for aa in AA if aa != code
]
else:
for chain in model.get_chains():
for residue in chain:
if pp.is_aa(residue):
code = pp.three_to_one(residue.get_resname())
position = residue.id[1]
chain_id = chain.id
prefix = code + chain_id + str(position)
mutations[prefix] = [
prefix + aa for aa in AA if aa != code
]
return mutations
def get_chain_to_valid_residues(structure, pdb_name=None):
"""Get tuples of chains and their valid residues."""
if pdb_name is None:
pdb_name = ()
else:
pdb_name = (pdb_name, )
chain_res = []
if type(structure) is Bio.PDB.Structure.Structure:
for model in structure:
for chain in model:
residues = [
res for res in chain
if poly.is_aa(res.get_resname(), standard=True)
and 'CA' in res
]
if len(residues) != 0:
chain_res.append(
(pdb_name + (str(model.serial_num), chain.get_id()),
residues))
else:
if 'atom_name' in structure.columns:
calphas = structure[structure['atom_name'] == 'CA']
else:
calphas = structure[structure['maestro_atom_name'] == ' CA ']
calphas = calphas[calphas['resname'] != 'UNK']
for (chain, chain_ca) in calphas.groupby(['model', 'chain']):
residues = [ca for idx, ca in chain_ca.iterrows()]
if len(residues) != 0:
chain_res.append((pdb_name + chain, residues))
return chain_res
def definePeptideChain(self): # find peptide chain if not stated in self.__table and fill self.__peptide
l = 'INFINITY' # with list of peptide residues if length is less than 30
if not self.__table['chain_antigen'][0]:
for i in self.__chains:
buf = len(PPBuilder().build_peptides(self.__struct[0][i])[0])
if (buf <= l):
l = buf
chid = i
self.__table.loc['chain_antigen', :] = chid
else:
chid = self.__table['chain_antigen'][0]
pp = list(self.__struct[0][chid])
if (len(pp) > 30):
line = self.__name + '\t;TOO MANY AMINO ACIDS (' + str(len(pp)) + ') TO BE A PEPTIDE :(\n'
self.printerr('definePeptideChain(): ' + line)
return 0
pep_res = []
for r in pp:
if (Polypeptide.is_aa(r.get_resname(), standard=True)):
pep_res.append(r)
self.__peptide = pep_res
self.__regions_res.update({'peptide':pep_res})
return 1
def aa_to_index(aa):
"""
:param aa: Three character amino acid name.
:returns: Integer index as per BioPython, unknown/non-standard amino acids return 20.
"""
if Polypeptide.is_aa(aa, standard=True):
return Polypeptide.three_to_index(aa)
else:
return 20
def get_chain_sequences(df):
"""Return list of tuples of (id, sequence) for different chains of monomers in a given dataframe."""
# Keep only CA of standard residues
df = df[df['name'] == 'CA'].drop_duplicates()
df = df[df['resname'].apply(lambda x: Poly.is_aa(x, standard=True))]
df['resname'] = df['resname'].apply(Poly.three_to_one)
chain_sequences = []
for c, chain in df.groupby(['ensemble', 'subunit', 'structure', 'model', 'chain']):
seq = ''.join(chain['resname'])
chain_sequences.append((tuple([str(x) for x in c]), seq))
return chain_sequences
def standard_residue_filter(df):
"""Filter out non-standard residues."""
residues = df[['structure', 'model', 'chain', 'residue', 'resname']] \
.drop_duplicates()
sel = residues['resname'].apply(lambda x: Poly.is_aa(x, standard=True))
residues['to_keep'] = sel
residues_to_keep = residues.set_index(
['structure', 'model', 'chain', 'residue', 'resname'])['to_keep']
to_keep = residues_to_keep.loc[df.set_index(
['structure', 'model', 'chain', 'residue', 'resname']).index]
return df[to_keep.values]
def get_all_chain_sequences_df(df):
"""Return list of tuples of (struct_name, chain_sequences) for sharded."""
all_chain_sequences = []
# Keep only CA of standard residues
df = df[df['name'] == 'CA'].drop_duplicates()
df = df[df['resname'].apply(lambda x: Poly.is_aa(x, standard=True))]
df['resname'] = df['resname'].apply(Poly.three_to_one)
for s, structure in df.groupby(['ensemble', 'subunit', 'structure']):
chain_sequences = []
for c, chain in structure.groupby(['model', 'chain']):
seq = ''.join(chain['resname'])
chain_sequences.append((c, seq))
all_chain_sequences.append((s, chain_sequences))
return all_chain_sequences
def write_FASTAs(PDB_ID, chains):
polypeptide_IDs = []
for chain_ID, residues in chains.items():
if residues and Polypeptide.is_aa(residues[0][0]):
polypeptide_ID = '{}_{}'.format(PDB_ID, chain_ID)
polypeptide_IDs.append(polypeptide_ID)
sequence = []
for resname, resseq, icode in residues:
try:
sequence.append(Polypeptide.three_to_one(resname))
except KeyError:
sequence.append('X')
with open('{}.fasta'.format(polypeptide_ID), mode='w') as f:
f.write('>{}\n'.format(polypeptide_ID))
f.write('{}\n'.format(''.join(sequence)))
return polypeptide_IDs
def get_bfactors( self, chain_id ):
'''
Input:
self: Use Biopython.PDB structure which has been stored in an object variable
chain_id : String (usually in ['A','B', 'C' ...]. The number of chains
depends on the specific protein and the resulting structure)
Return:
Return the B-Factors for all residues in a chain of a Biopython.PDB structure.
The B-Factors describe the mobility of an atom or a residue.
In a Biopython.PDB structure B-Factors are given for each atom in a residue.
Calculate the mean B-Factor for a residue by averaging over the B-Factor
of all atoms in a residue.
Sometimes B-Factors are not available for a certain residue;
(e.g. the residue was not resolved); insert np.nan for those cases.
Finally normalize your B-Factors using Standard scores (zero mean, unit variance).
You have to use np.nanmean, np.nanvar etc. if you have nan values in your array.
The returned data structure has to be a numpy array rounded again to integer.
'''
chain = self.structure.child_list[0].child_dict[chain_id]
residues1 = chain.get_list()
residues = []
#remove residues that are not AAs
for res_nr in range(len(residues1)):
if Polypeptide.is_aa(residues1[res_nr]):
residues.append(residues1[res_nr])
length = len(residues)
b_factors = np.zeros(length, dtype=np.float32)
#calculate bfactor average per residue
tmp_factors = np.zeros(length, dtype=np.float32)
for res_nr in range(length):
atoms = residues[res_nr].get_list()
bfactor = 0
atom_count = 0
for ato in atoms:
bfactor += ato.bfactor
atom_count += 1
bfactor /= atom_count
tmp_factors[res_nr] = bfactor
#normalize
mean = np.nanmean(tmp_factors)
vari = np.nanstd(tmp_factors)
for foo in range(length):
b_factors[foo] = (tmp_factors[foo]-mean)/vari
return b_factors.astype( np.int ) # return rounded (integer) values
def get_sequence( self, chain_id ):
'''
Input:
self: Use Biopython.PDB structure which has been stored in an object variable
chain_id : String (usually in ['A','B', 'C' ...]. The number of chains
depends on the specific protein and the resulting structure)
Return:
Return the amino acid sequence (single-letter alphabet!) of a given chain (chain_id)
in a Biopython.PDB structure as a string.
'''
chain = self.structure.child_list[0].child_dict[chain_id]
sequence = ""
for residue in chain:
if Polypeptide.is_aa(residue):
long_name = residue.get_resname()
sequence += Polypeptide.three_to_one(long_name)
return sequence
def clean_pdb(structure, pdb_name, clean_dir):
'''
Function to select and write pdb with only aminoacids
Invokes SelectAA class constructed with Bio.PDB.select
Called by: clean_pdb_files()
clean_and_sort()
'''
reslist = []
clean_name = clean_dir + pdb_name + '.clean.pdb'
for res in structure.get_residues():
if bpp_poly.is_aa(res.get_resname(), standard=True):
reslist.append(res.resname)
if len(reslist) > 30:
io.set_structure(structure)
io.save(clean_name, SelectAA())
return True
else:
return False
def standard_residue_filter(df):
"""
Filter out non-standard residues.
:param df: dataframe to filter against.
:type df: atoms dataframe.
:return: same dataframe, but with only with atoms corresponding to standard residues left.
:rtype: atoms dataframe.
"""
residues = df[['structure', 'model', 'chain', 'residue', 'resname']] \
.drop_duplicates()
sel = residues['resname'].apply(
lambda x: Poly.is_aa(x, standard=True))
residues['to_keep'] = sel
residues_to_keep = residues.set_index(
['structure', 'model', 'chain', 'residue', 'resname'])['to_keep']
to_keep = residues_to_keep.loc[df.set_index(
['structure', 'model', 'chain', 'residue', 'resname']).index]
return df[to_keep.values]
def extract_seqs(structure, defmodel):
'''
Uses Biopython to count the numer of chains and to extract the
each chain's sequence as a list of sequences.
Called by: clean_and_sort()
'''
nchains = 0
for model in structure:
if model.id == defmodel:
seqs = []
chain_ids = []
for chain in model:
nchains += 1
seqlist = []
for residue in chain:
if bpp_poly.is_aa(residue.get_resname(), standard=True):
seqlist.append(
bpp_poly.three_to_one(residue.get_resname()))
else:
seqlist.append('X')
seq = str("".join(seqlist))
seqs.append(seq)
chain_ids.append(chain.id)
return nchains, seqs, chain_ids
def get_contact_map( self, chain_id ):
'''
Input:
self: Use Biopython.PDB structure which has been stored in an object variable
chain_id : String (usually in ['A','B', 'C' ...]. The number of chains
depends on the specific protein and the resulting structure)
Return:
Return a complete contact map (see description in exercise sheet)
for a given chain in a Biopython.PDB structure as numpy array.
The values in the matrix describe the c-alpha distance between all residues
in a chain of a Biopython.PDB structure.
Only integer values of the distance have to be given (see below).
'''
length = len(self.get_sequence(chain_id))
contact_map = np.zeros( (length,length), dtype=np.float32 )
aa = []
for residue in self.structure.child_list[0].child_dict[chain_id]:
if Polypeptide.is_aa(residue):
aa.append(residue)
for foo in range(0,length):
for bar in range(0,length):
contact_map[foo][bar] = self.get_residue_distance(aa[foo], aa[bar])
return contact_map.astype( np.int ) # return rounded (integer) values
def _aa_mask(self):
poly = Polypeptide.Polypeptide(self._bio_chain)
aa_mask = [Polypeptide.is_aa(r) for r in poly]
return aa_mask
def get_structure_seqrecords(model):
"""Get a dictionary of a PDB file's sequences.
Special cases include:
- Insertion codes. In the case of residue numbers like "15A", "15B", both residues are written out. Example: 9LPR
- HETATMs. Currently written as an "X", or unknown amino acid.
Args:
model: Biopython Model object of a Structure
Returns:
list: List of SeqRecords
"""
structure_seq_records = []
# Loop over each chain of the PDB
for chain in model:
tracker = 0
chain_seq = ''
chain_resnums = []
# Loop over the residues
for res in chain.get_residues():
# NOTE: you can get the residue number too
res_id = res.id
res_num = res_id[1]
res_icode = res_id[2]
# Double check if the residue name is a standard residue
# If it is not a standard residue (ie. selenomethionine),
# it will be filled in with an X on the next iteration)
if Polypeptide.is_aa(res, standard=True):
end_tracker = res_num
res_aa_one = Polypeptide.three_to_one(res.get_resname())
# Tracker to fill in X's
if end_tracker != (tracker + 1):
if res_icode != ' ':
chain_seq += res_aa_one
chain_resnums.append(res_num)
tracker = end_tracker + 1
continue
else:
multiplier = (end_tracker - tracker - 1)
chain_seq += 'X' * multiplier
# Residue numbers for unresolved or nonstandard residues are Infinite
chain_resnums.extend([float("Inf")] * multiplier)
chain_seq += res_aa_one
chain_resnums.append(res_num)
tracker = end_tracker
else:
continue
chain_seq_record = SeqRecord(Seq(chain_seq, IUPAC.protein),
id=chain.get_id())
chain_seq_record.letter_annotations[
'structure_resnums'] = chain_resnums
structure_seq_records.append(chain_seq_record)
return structure_seq_records
def build_matrix(
path: str,
filename: str,
truncate_log: Union[tqdm.tqdm, None] = None) -> BuildMatrixDict:
"""Build the input matrix for one protein.
Args:
path: path of the pdb file.
filename: name of the file (without extension).
truncate_log: tqdm logger
Returns:
Build matrix dictionary
"""
PROTEIN_SEQ_MAX_LEN = 4000
protein_matrix = [[0 for x in range(PROTEIN_SEQ_MAX_LEN)]
for y in range(10)]
protein_structure = PDBParser().get_structure(filename, path)
protein_model = list(protein_structure.get_models())
protein_chains = list(protein_model[0].get_chains())
col = 0
try:
for chain in protein_chains:
protein_residues = list(chain.get_residues())
for residue in protein_residues:
if Polypeptide.is_aa(residue.get_resname(), standard=True):
atoms = list(residue.get_atoms())
x = []
y = []
z = []
for atom in atoms:
vec = atom.get_vector()
x.append(vec.__getitem__(0))
y.append(vec.__getitem__(1))
z.append(vec.__getitem__(2))
# calculate position of residue
x = round(mean(x))
y = round(mean(y))
z = round(mean(z))
# one letter code
code = Polypeptide.three_to_one(residue.get_resname())
aa = amino_acid[code]
protein_matrix[0][col] = aa["code"]
protein_matrix[1][col] = x
protein_matrix[2][col] = y
protein_matrix[3][col] = z
protein_matrix[4][col] = aa["hydropathy"]
protein_matrix[5][col] = aa["hydropathy_index"]
protein_matrix[6][col] = aa["acidity_basicity"]
protein_matrix[7][col] = aa["mass"]
protein_matrix[8][col] = aa["isoelectric_point"]
protein_matrix[9][col] = aa["charge"]
# Even if the current residue is not amino acid we increase the col.
# 0 is save at this position if it is not an amino acid.
col = col + 1
except IndexError:
if truncate_log is not None:
truncate_log.set_description_str(
f"Protein {filename} is truncated.")
# Prepare dict so it can be load to vaex dataframe
dic: BuildMatrixDict = {
"seq": [[]],
"x_pos": [[]],
"y_pos": [[]],
"z_pos": [[]],
"hydropathy": [[]],
"hydropathy_index": [[]],
"acidity_basicity": [[]],
"mass": [[]],
"isoelectric_point": [[]],
"charge": [[]],
}
for i in range(10):
dic[col_name[i]] = pyarrow.array(
[[protein_matrix[i][x] for x in range(PROTEIN_SEQ_MAX_LEN)]])
return dic
def get_structure_seqs(pdb_file, file_type):
"""Get a dictionary of a PDB file's sequences.
Special cases include:
- Insertion codes. In the case of residue numbers like "15A", "15B", both residues are written out. Example: 9LPR
- HETATMs. Currently written as an "X", or unknown amino acid.
Args:
pdb_file: Path to PDB file
Returns:
dict: Dictionary of:
{chain_id: sequence}
"""
# TODO: Please check out capitalization of chain IDs in mmcif files. example: 5afi - chain "l" is present but
# it seems like biopython capitalizes it to chain L
# Get the first model
my_structure = StructureIO(pdb_file)
model = my_structure.first_model
structure_seqs = {}
# Loop over each chain of the PDB
for chain in model:
chain_seq = ''
tracker = 0
# Loop over the residues
for res in chain.get_residues():
# NOTE: you can get the residue number too
# res_num = res.id[1]
# Double check if the residue name is a standard residue
# If it is not a standard residue (ie. selenomethionine),
# it will be filled in with an X on the next iteration)
if Polypeptide.is_aa(res, standard=True):
full_id = res.get_full_id()
end_tracker = full_id[3][1]
i_code = full_id[3][2]
aa = Polypeptide.three_to_one(res.get_resname())
# Tracker to fill in X's
if end_tracker != (tracker + 1):
if i_code != ' ':
chain_seq += aa
tracker = end_tracker + 1
continue
else:
chain_seq += 'X' * (end_tracker - tracker - 1)
chain_seq += aa
tracker = end_tracker
else:
continue
structure_seqs[chain.get_id()] = chain_seq
return structure_seqs
def get_knots(pdb, cutoff, cluster_cutoff, genpdb, verbosity):
'''
Main routine, uses biopython and pandas to detect knots and cluster them
through the implementation of the average linkage algorithm
Called by: main()
'''
if (pdb.endswith(".ent") or pdb.endswith(".pdb") or pdb.endswith(".ent.gz")
or pdb.endswith(".pdb1") or pdb.endswith(".pdb1.gz")
or pdb.endswith(".pdb.gz")) and not pdb.startswith('CONTACTS-'):
pdb_name, structure, nchains = strtools.parse_pdb_structure(pdb)
print(str('\n' + clrs['p'] + pdb + clrs['n']))
with open('KnotScope.log', 'a') as log:
log.write(str('\n[STRUCTURE],' + pdb + '\n'))
mainchain = [
atom for atom in bpp.Selection.unfold_entities(structure[0], 'A')
if bpp_poly.is_aa(atom.get_parent(), standard=True) and (
atom.id == 'CA')
] # or atom.id == 'N' or atom.id == 'O')]
contacts = []
core = []
for atom in mainchain:
distances = []
ns = bpp.NeighborSearch(mainchain)
center = atom.get_coord()
neighbors = [
neighbor for neighbor in ns.search(center, cutoff)
if (neighbor.get_parent().id[1] -
atom.get_parent().id[1]) > abs(3)
]
if neighbors:
for neighbor in neighbors:
d = neighbor - atom
distances.append(d)
if d <= cutoff:
printv(
clrs['y'] + 'Unlikely proximity' + clrs['n'] +
' between residues ' + clrs['y'] +
str(atom.get_parent().id[1]) + clrs['n'] +
' and ' + clrs['y'] +
str(neighbor.get_parent().id[1]) + clrs['n'] + '!',
verbosity)
printv(str(d), verbosity)
with open('KnotScope.log', 'a') as log:
log.write('[CLASH],' +
str(atom.get_parent().id[1]) + ',' +
str(neighbor.get_parent().id[1]) + ',' +
str(d) + '\n')
contacts.append(neighbor.get_parent())
contacts.append(atom.get_parent())
if atom not in core:
core.append(atom)
if neighbor not in core:
core.append(neighbor)
# Save contacts to pdb file if they exist
if contacts and genpdb:
io.set_structure(structure)
io.save('CONTACTS-' + pdb, strtools.SelectResidues(contacts))
# Start cluster analysis to separate knots
pairwisedist = []
# Measure pairwise distances of every CA involved in knots and record in vertical list
if len(core) > 1:
for a, b in it.combinations(core, 2):
d = a - b
pairwisedist.append(
[a.get_parent().id[1],
b.get_parent().id[1], d])
# Add values for diagonal
for entry in range(len(core)):
line = make_diagonal(core, entry)
pairwisedist.append([line[0], line[1], line[2]])
# Create pandas dataframe, make it a square and symmetric matrix
df = pd.DataFrame(pairwisedist, index=None, columns=None)
df = pd.crosstab(index=df[0],
columns=df[1],
values=df[2],
aggfunc='sum',
dropna=True).fillna(0)
df = df + df.T
# Start average linkage algorithm
reslist = list(df.columns)
clusters = []
row_index = -1
col_index = -1
array = []
for n in range(df.shape[0]):
array.append(n)
clusters.append(array.copy())
for k in range(1, df.shape[0]):
min_val = sys.maxsize
for i in range(0, df.shape[0]):
for j in range(0, df.shape[1]):
#print(str(df.iloc[i,j]))
if type(df.iloc[i, j]) != str:
if (df.iloc[i, j] <= min_val):
min_val = df.iloc[i, j]
row_index = i
col_index = j
for i in range(0, df.shape[0]):
if (i != col_index and i != row_index):
temp = (df.iloc[col_index, i] +
df.iloc[row_index, i]) / 2
df.iloc[col_index, i] = temp
df.iloc[i, col_index] = temp
for i in range(0, df.shape[0]):
df.iloc[row_index, i] = sys.maxsize
df.iloc[i, row_index] = sys.maxsize
minimum = min(row_index, col_index)
maximum = max(row_index, col_index)
for n in range(len(array)):
if (array[n] == maximum):
array[n] = minimum
clusters.append(array.copy())
# Stop iterations when minimum pairwise distance in the matrix is greater than 22
if min_val > cluster_cutoff:
break
# Get the clusters from last iteration and 'count' elements
clustered_res = clusters[-1]
counter = collections.Counter(clustered_res)
# Combine residue and cluster information and print them user-friendly
clusterdict = dict(zip(reslist, clustered_res))
print(clrs['y'] + '\nLikely ' + str(len(set(clusters[-1]))) +
' knot(s) found in structure under chosen criteria...' +
clrs['n'])
n = 0
k_lengths = []
for cl in set(clusterdict.values()):
n += 1
cluster_residues = []
print('\nKnot ' + clrs['y'] + str(n) + clrs['n'] +
' (Cluster id: ' + str(cl) + ') involves ' + clrs['y'] +
str(list(counter.values())[n - 1]) + clrs['n'] +
' residues:')
for res in clusterdict:
if clusterdict[res] == cl:
cluster_residues.append(res)
k_lengths.append(len(cluster_residues))
print(clrs['y'] + ', '.join([str(a)
for a in cluster_residues]) +
clrs['n'])
with open('KnotScope.log', 'a') as log:
log.write('[K' + str(n) + '-RES],' +
','.join([str(a)
for a in cluster_residues]) + '\n')
log.write('[K' + str(n) + '-LEN],' +
str(len(cluster_residues)) + '\n')
if clusterdict:
nknots = len(set(list(clusterdict.values())))
else:
nknots = 0
maxklength = max(k_lengths)
with open('KnotScope.log', 'a') as log:
log.write('[SUM],str,' + pdb + ',ca_clash,' + str(len(core)) +
',nknots,' + str(nknots) + ',maxklength,' +
str(maxklength) + '\n')
return clusterdict, nknots, maxklength
else:
print(clrs['g'] + 'No CA distances under ' + str(cutoff) +
' angstrons found' + clrs['n'] + '!\n')
with open('KnotScope.log', 'a') as log:
log.write('[SUM],str,' + pdb +
',ca_clash,0,nknots,0,maxklength,0\n')
del pdb_name, structure, nchains, contacts
elif pdb.startswith('CONTACTS-'):
pass
else:
print(clrs['y'] + pdb + clrs['n'] +
' not a pdb-related structure format.' + clrs['r'] +
' SKIPPING!' + clrs['n'])
def aa_to_index(aa):
if Polypeptide.is_aa(aa, standard=True):
return Polypeptide.three_to_index(aa)
else:
return 20
def accept_residue(self, residue):
if bpp_poly.is_aa(residue.get_resname(), standard=True):
return 1
else:
return 0