def gather_pdb(k):
# For each PDB in kinDB.xml, download the PDB file and SIFTS residue-mapping .xml file if they are not already present
pdb_nodes = kinDB[k].findall('pk_pdb')
uniprot_sequence = kinDB[k].findtext('uniprot/sequence').strip()
uniprot_sequence = ''.join(uniprot_sequence.split('\n'))
uniprotAC = kinDB[k].find('uniprot').get('AC')
entry_name = kinDB[k].find('uniprot').get('entry_name')
#if uniprotAC != 'P00533':
# return
kinase_results = []
for pdb_node in pdb_nodes:
pdbid = pdb_node.get('id')
#if pdbid != '2ITN':
# continue
# Download PDB file if necessary
local_pdb_file_path = os.path.join(local_pdb_path, pdbid+'.pdb')
if os.path.exists(local_pdb_file_path):
pass
else:
print 'Downloading PDB file and saving as:', local_pdb_file_path
page = retrieve_pdb(pdbid, compressed='yes')
with gzip.open(local_pdb_file_path, 'wb') as local_pdb_file:
local_pdb_file.write(page + '\n')
# Download SIFTS file if necessary
local_sifts_file_path = os.path.join(local_sifts_path, pdbid+'.xml.gz')
if os.path.exists(local_sifts_file_path):
pass
else:
print 'Downloading SIFTS file and saving as (compressed):', local_sifts_file_path
page = retrieve_sifts(pdbid)
with gzip.open(local_sifts_file_path, 'wb') as local_sifts_file:
local_sifts_file.write(page + '\n')
# Parse the sifts XML document
with gzip.open(local_sifts_file_path,'rb') as local_sifts_file:
sifts = etree.parse(local_sifts_file, parser).getroot()
# From the PDB file, get the EXPRESSION_SYSTEM and related fields
expression_data = dict()
with open(local_pdb_file_path,'r') as local_pdb_file:
for line in local_pdb_file.readlines():
regex_search = re.search('EXPRESSION_SYSTEM.*:', line)
if regex_search != None:
key = line[regex_search.start() : regex_search.end() - 1]
data = line[regex_search.end() + 1 : ].strip()
if data[-1] == ';':
data = data[:-1]
expression_data[key] = data
# Get the chains to be searched from kinDB
kinDB_chain_nodes = pdb_node.findall('chain')
pdb_results = []
for chain_node in kinDB_chain_nodes:
DELETE_ME = False
chainid = chain_node.get('id')
if verbose: print entry_name, uniprotAC, pdbid, chainid
# First check whether the first residue with matching chainid and a UniProt crossref has the same UniProt AC as was picked up from UniProt (by gather-uniprot.py).
# 3O50 and 3O51 are picked up by gather-uniprot.py from uniprot AC O14965. But these have uniprot AC B4DX16 in the sifts .xml files, which is a TrEMBL entry. Sequences are almost identical except for deletion of ~70 residues prior to PK domain of B4DX16. This means that experimental_sequence_aln and related sequences are not added by gather-pdb.py. Need to sort out a special case for these pdbs. Should check for similar cases in other kinases.
# 3O50 and 3O51 can be ignored. (Plenty of other PDBs for that protein)
# 3OG7 is picked up from uniprot AC P15056, but the PDB entry links to Q5IBP5 - this is the AKAP9-BRAF fusion protein.
# XXX TODO XXX 3OG7 will be ignored for now, but at some point should make separate entries for fusion proteins, and add the PDB files accordingly.
if verbose: print sifts
first_matching_uniprot_resi = sifts.find('entity[@type="protein"]/segment/listResidue/residue/crossRefDb[@dbSource="PDB"][@dbChainId="%s"]/../crossRefDb[@dbSource="UniProt"]' % chainid)
sifts_uniprotAC = first_matching_uniprot_resi.get('dbAccessionId')
if uniprotAC != sifts_uniprotAC:
print 'PDB %s chain %s picked up from UniProt entry %s %s. Non-matching UniProtAC in sifts: %s. This pk_pdb entry will be deleted when outputting %s' % (pdbid, chainid, entry_name, uniprotAC, sifts_uniprotAC, okinDB_path)
DELETE_ME = True
#
#
# TODO check if there are any PDBs where two proteins share the same chainid (I seem to remember that there are - check previous scripts)
#
#
# Now extract the sequence data
# These are the sifts residues which include a PDB crossref with matching chainid
chain_residues = sifts.findall('entity[@type="protein"]/segment/listResidue/residue/crossRefDb[@dbSource="PDB"][@dbChainId="%s"]/..' % chainid)
experimental_sequence = ''
experimental_sequence_pdb_resids = []
experimental_sequence_uniprot_res_indices = []
observed_sequence_aln_exp = ''
experimental_sequence_aln = ['-'] * len(uniprot_sequence) # This will contain the alignment of the experimental sequence against the full UniProt sequence. Conflicting residues will be added if they are contiguous with non-conflicting segments. NOTE: this is no longer added to the database.
experimental_sequence_aln_conflicts = ['-'] * len(uniprot_sequence) # Same, but conflicting residues are added as lower case
observed_sequence_aln = ['-'] * len(uniprot_sequence) # This will contain the alignment of the observed sequence against the full UniProt sequence. Conflicting residues will be ignored.
ss_aln = ['-'] * len(uniprot_sequence) # This will contain the alignment of the secondary structure codes against the full UniProt sequence. Conflicting residues will be ignored.
n_crossref_uniprot_matches = 0
for r in chain_residues:
residue_details = r.findall('residueDetail')
residue_detail_texts = [ detail.text.strip() for detail in residue_details ] # list of strings
ss = r.findtext('residueDetail[@property="codeSecondaryStructure"]')
resname = r.attrib['dbResName']
if resname == None:
print 'ERROR: UniProt crossref not found for conflicting residue!', k, pdbid, chainid, r.attrib
raise Exception
try:
# Note that this BioPython dict converts a modified aa to the single-letter code of its unmodified parent (e.g. "TPO":"T")
single_letter = to_one_letter_code[ resname ]
except KeyError:
if resname == 'ACE': # Just ignore N-terminal ACE
continue
elif resname == 'CAS': # S-(dimethylarsenic)cysteine
single_letter = 'C'
elif resname == 'MHO': # S-oxymethionine
single_letter = 'M'
elif resname == 'LGY': # 3NX8. (E)-N-(4-oxobutylidene)lysine
single_letter = 'K'
elif resname == 'AME': # N-acetylmethionine
single_letter = 'M'
elif resname == 'NMM': # 3KB7
single_letter = 'R'
elif resname == 'OCY': # 2R9S
single_letter = 'C'
elif resname == 'CY0': # 2J5E
single_letter = 'C'
elif resname == 'CY7': # 2JIV
single_letter = 'C'
else:
print 'KeyError: Problem converting resname', resname, 'to single letter code.', k, pdbid, chainid, r.attrib
raise KeyError
# Add residue to experimental_sequence
experimental_sequence += single_letter
# Also save the pdb resids, which we will use later
pdb_resid = r.find('crossRefDb[@dbSource="PDB"]').attrib['dbResNum']
# Some pdb resids are e.g. '464A'
if pdb_resid.isdigit() == False:
if pdbid in ['1O6L','2JDO','2JDR','2UW9','2X39','2XH5']: # These pdbs include three residues with pdb resids 464A, 464B, 464C, (all with UniProt crossrefs) then continues from 465. We will change this so that the pdb resids continue to iterate
corrected_pdb_resids = {'464A':465, '464B':466, '464C':467}
if pdb_resid in corrected_pdb_resids.keys():
pdb_resid = corrected_pdb_resids[pdb_resid]
elif int(pdb_resid[0:3]) > 464:
pdb_resid = int(pdb_resid) + 3
# Otherwise just extract the number (this will also detect negative numbers)
else:
pdb_resid = ''.join([char for char in pdb_resid if (char.isdigit() or char == '-')])
try:
experimental_sequence_pdb_resids.append( int(pdb_resid) )
except:
print 'Problem converting pdb_resid into int.' , uniprotAC, pdbid, chainid, pdb_resid
raise Exception
# Also add residue to experimental_sequence_aln. Residues which do not match the uniprot sequence (and thus do not have a uniprot crossref) will be added later
crossref_uniprot = r.find('crossRefDb[@dbSource="UniProt"][@dbAccessionId="%s"]' % uniprotAC)
if crossref_uniprot != None:
n_crossref_uniprot_matches += 1
index = int(crossref_uniprot.attrib['dbResNum']) - 1
experimental_sequence_aln[index] = single_letter
if 'Conflict' in residue_detail_texts:
experimental_sequence_aln_conflicts[index] = single_letter.lower()
else:
experimental_sequence_aln_conflicts[index] = single_letter
experimental_sequence_uniprot_res_indices.append(index)
# Add residue to observed_sequence_aln if it is observed and is not a conflict
if 'Not_Observed' not in residue_detail_texts and 'Conflict' not in residue_detail_texts:
observed_sequence_aln[index] = single_letter
if ss != None:
ss_aln[index] = ss
else:
experimental_sequence_uniprot_res_indices.append(None)
pass
# Add residue to observed_sequence_aln_exp if it is observed, otherwise '-'
if 'Not_Observed' in residue_detail_texts:
observed_sequence_aln_exp += '-'
else:
observed_sequence_aln_exp += single_letter
# Now check whether the number of non-observed residues is more than 90% of the experimental sequence length
n_unobserved_residues = observed_sequence_aln_exp.count('-')
if ( float(n_unobserved_residues) / float(len(experimental_sequence)) ) > 0.9:
DELETE_ME = True
# ======
# Now we add the residues which do not have a uniprot crossref
# ======
#print k, uniprotAC, pdbid, chainid
#print experimental_sequence
#print ''.join(experimental_sequence_aln_conflicts)
i = 0
# But first we have to deal with cases where residues have been added at the N-terminus which extend before the start of the uniprot sequence. The excess residues will be ignored.
# Get the uniprot residue index of the first residue with a uniprot crossref
for s in range(len(experimental_sequence_uniprot_res_indices)):
UP_res_index = experimental_sequence_uniprot_res_indices[s]
if UP_res_index != None:
first_exp_seq_uniprot_res_index = UP_res_index
# And the corresponding pdb resid
corresponding_pdb_resid = experimental_sequence_pdb_resids[s]
exp_seq_first_uniprot_res_index = s
break
# And get the pdb resid of the first residue in the experimental sequence
for s in experimental_sequence_pdb_resids:
if s != None:
first_exp_seq_pdb_resid = s
break
ignore_excess_Nterm_residues_flag = False
# If the experimental sequence includes the first residue of the full uniprot sequence
if first_exp_seq_uniprot_res_index == 0:
# And if the value of the first pdb resid is lower than that of the pdb resid corresponding to the first uniprot residue
if first_exp_seq_pdb_resid < corresponding_pdb_resid:
# Then we will ignore the excess residues
ignore_excess_Nterm_residues_flag = True
# Now iterate through the residues in the experimental sequence and add residues which do not have a uniprot crossref, but are contiguous in terms of PDB numbering
while i < len(experimental_sequence):
resname_i = experimental_sequence[i]
uniprot_res_index_i = experimental_sequence_uniprot_res_indices[i]
pdb_resid_i = experimental_sequence_pdb_resids[i]
if (ignore_excess_Nterm_residues_flag == True) and (pdb_resid_i < corresponding_pdb_resid):
pass # we ignore these residues
# If this residue does not have a uniprot crossref
elif uniprot_res_index_i == None:
# Start a list of residues with no uniprot crossref
contiguous_noUP_residues = [ resname_i ]
# Then check the next residue
j = i + 1
while j < len(experimental_sequence):
resname_j = experimental_sequence[j]
uniprot_res_index_j = experimental_sequence_uniprot_res_indices[j]
pdb_resid_j = experimental_sequence_pdb_resids[j]
#print 'len, i, j:', len(experimental_sequence), i, j, pdb_resid_i, pdb_resid_j, contiguous_noUP_residues
# If this residue also has no uniprot crossref, and is contiguous in terms of pdb resnum, then add it to the list, and move on to the next one
if (uniprot_res_index_j == None) and ((pdb_resid_j - pdb_resid_i) == (j-i)):
#print 'adding to list:', j, resname_j
contiguous_noUP_residues.append( resname_j )
pass
# If this residue does have a uniprot crossref, and if it is contiguous in terms of pdb resnum, then we add the list of residues without uniprot crossrefs at this position
elif (uniprot_res_index_j != None) and ((pdb_resid_j - pdb_resid_i) == (j-i)):
#print 'adding to sequence_aln:', j
experimental_sequence_aln[ (uniprot_res_index_j - j) : uniprot_res_index_j ] = contiguous_noUP_residues
experimental_sequence_aln_conflicts[ (uniprot_res_index_j - j) : uniprot_res_index_j ] = list(''.join(contiguous_noUP_residues).lower())
i = j
break
# If this residue is not contiguous in terms of pdb resnum, go back and check if the first of contiguous_noUP_residues is pdb-contiguous with the previous residue - if so, add contiguous_noUP_residues
elif (pdb_resid_j - pdb_resid_i) != (j-i):
#print 'checking backwards:', j
if (pdb_resid_i - experimental_sequence_pdb_resids[i-1]) == 1:
last_uniprot_res_index = experimental_sequence_uniprot_res_indices[i-1]
experimental_sequence_aln[ last_uniprot_res_index + 1 : last_uniprot_res_index + 1 + (j-i)] = contiguous_noUP_residues
experimental_sequence_aln_conflicts[ last_uniprot_res_index + 1 : last_uniprot_res_index + 1 + (j-i)] = list(''.join(contiguous_noUP_residues).lower())
i = j - 1
break
# If we have reached the end of experimental_sequence, go back and check if the first of contiguous_noUP_residues is pdb-contiguous with the previous residue - if so, add contiguous_noUP_residues
if j == len(experimental_sequence) - 1:
#print 'THIS IS THE END', len(experimental_sequence), i, j, pdb_resid_i, experimental_sequence_pdb_resids[i], experimental_sequence_pdb_resids[i-1], contiguous_noUP_residues
#print experimental_sequence_pdb_resids
if (pdb_resid_i - experimental_sequence_pdb_resids[i-1]) == 1:
last_uniprot_res_index = experimental_sequence_uniprot_res_indices[i-1]
experimental_sequence_aln[ last_uniprot_res_index + 1 : last_uniprot_res_index + 2 + (j-i)] = contiguous_noUP_residues
experimental_sequence_aln_conflicts[ last_uniprot_res_index + 1 : last_uniprot_res_index + 2 + (j-i)] = list(''.join(contiguous_noUP_residues).lower())
i = j
break
j += 1
i += 1
# In cases such as 3LAU and 1O6L, additional sequence at end makes experimental_sequence_aln longer than uniprot_sequence by 1
if len(experimental_sequence_aln) != len(uniprot_sequence):
experimental_sequence_aln = experimental_sequence_aln[0:len(uniprot_sequence)]
experimental_sequence_aln_conflicts = experimental_sequence_aln_conflicts[0:len(uniprot_sequence)]
#print ''.join(experimental_sequence_aln_conflicts)
# Now add the various sequence data to kinDB
experimental_sequence_aln = ''.join(experimental_sequence_aln)
experimental_sequence_aln_conflicts = ''.join(experimental_sequence_aln_conflicts)
observed_sequence_aln = ''.join(observed_sequence_aln)
ss_aln = ''.join(ss_aln)
#exp = etree.SubElement(chain_node,'experimental')
#etree.SubElement(exp,'sequence').text = seqwrap(experimental_sequence)
#etree.SubElement(exp,'sequence_aln').text = seqwrap(experimental_sequence_aln)
#etree.SubElement(exp,'sequence_aln_conflicts').text = seqwrap(experimental_sequence_aln_conflicts)
#obs = etree.SubElement(chain_node,'observed')
#etree.SubElement(obs,'sequence').text = seqwrap(observed_sequence)
result = [experimental_sequence, experimental_sequence_aln, experimental_sequence_aln_conflicts, observed_sequence_aln_exp, observed_sequence_aln, ss_aln, DELETE_ME]
pdb_results.append(result)
pdb_results.append(expression_data)
kinase_results.append(pdb_results)
return kinase_results
# =================================
print '\n= Require %s .pdb and sifts .xml files (will be downloaded if they don\'t exist)=\n' % len(
pdbids)
if not os.path.exists(structures_dir):
print 'Making directory', structures_dir
os.mkdir(structures_dir)
for t in template_ids:
# Get the pdbid by splitting the template_id by '_'
pdbid = t.split('_')[4]
pdb_filename = os.path.join(pdb_dir, pdbid + '.pdb')
pdbgz_filename = os.path.join(pdb_dir, pdbid + '.pdb.gz')
sifts_filename = os.path.join(sifts_dir, pdbid + '.xml.gz')
if not os.path.exists(pdb_filename):
print 'Downloading PDB file for:', pdbid
# Download compressed file, then uncompress
pdbgz_page = retrieve_pdb(pdbid, compressed='yes')
with open(pdbgz_filename, 'w') as pdbgz_file:
pdbgz_file.write(pdbgz_page)
with gzip.open(pdbgz_filename, 'rb') as pdbgz_file_decoded:
with open(pdb_filename, 'w') as pdb_filename:
pdb_filename.writelines(pdbgz_file_decoded)
os.remove(pdbgz_filename)
if not os.path.exists(sifts_filename):
print 'Downloading sifts file for:', pdbid
sifts_page = retrieve_sifts(pdbid)
with gzip.open(sifts_filename, 'wb') as sifts_file:
sifts_file.write(sifts_page)
if not os.path.exists(template_structures_dir):
os.mkdir(template_structures_dir)
# Download .pdb and sifts .xml files
# =================================
print '\n= Require %s .pdb and sifts .xml files (will be downloaded if they don\'t exist)=\n' % len(pdbids)
if not os.path.exists(structures_dir):
print 'Making directory', structures_dir
os.mkdir(structures_dir)
for t in template_ids:
# Get the pdbid by splitting the template_id by '_'
pdbid = t.split('_')[4]
pdb_filename = os.path.join(pdb_dir, pdbid + '.pdb')
pdbgz_filename = os.path.join(pdb_dir, pdbid + '.pdb.gz')
sifts_filename = os.path.join(sifts_dir, pdbid + '.xml.gz')
if not os.path.exists(pdb_filename):
print 'Downloading PDB file for:', pdbid
# Download compressed file, then uncompress
pdbgz_page = retrieve_pdb(pdbid, compressed='yes')
with open(pdbgz_filename, 'w') as pdbgz_file:
pdbgz_file.write(pdbgz_page)
with gzip.open(pdbgz_filename, 'rb') as pdbgz_file_decoded:
with open(pdb_filename, 'w') as pdb_filename:
pdb_filename.writelines(pdbgz_file_decoded)
os.remove(pdbgz_filename)
if not os.path.exists(sifts_filename):
print 'Downloading sifts file for:', pdbid
sifts_page = retrieve_sifts(pdbid)
with gzip.open(sifts_filename, 'wb') as sifts_file:
sifts_file.write(sifts_page)
if not os.path.exists(template_structures_dir):
os.mkdir(template_structures_dir)