def match_renumber(self, reference_pdb):
""" Match the chains and renumber the structures according to a reference PDB """
clustalo_exe = ini.get('third_party', 'clustalo_exe')
if not shutil.which(clustalo_exe):
print('\n+ ClustalO not correctly configured in haddock3.ini')
print('+ WARNING: matching not possible!')
return False
else:
print('\n+ Running automated chain matching and renumbering')
print('+ WARNING: Use with caution, some residues could be deleted')
if reference_pdb == 'lowest':
reference_pdb, reference_score = self.fetch_lowest()
if ' ' not in PDB.identify_chains(reference_pdb):
reference_pdb = PDB.fix_id(reference_pdb, priority='chain')
if ' ' not in PDB.identify_chainseg(reference_pdb):
reference_pdb = PDB.fix_id(reference_pdb, priority='seg')
reference_seq_dic = PDB.load_seq(reference_pdb)
reference_chains = PDB.identify_chains(reference_pdb)
reference_chains.sort()
pdb_list = list(self.structure_dic.keys())
pdb_list.sort()
for pdb in pdb_list:
# match the chains with sequence alignment
target_seq_dic = PDB.load_seq(pdb)
pdb = PDB.fix_id(pdb, priority='seg')
# Get what chains are present in the target
target_chains = PDB.identify_chains(pdb)
target_chains.sort()
# Do a combinatorial alignment to check which chains match better
identity_dic = {}
for ref_chain, target_chain in itertools.product(reference_chains, target_chains):
ref_seq = ''.join(list(reference_seq_dic[ref_chain].values()))
target_seq = ''.join(list(target_seq_dic[target_chain].values()))
open('seq.fasta', 'w').write(f'>ref\n{ref_seq}\n>target\n{target_seq}\n')
cmd = f'{clustalo_exe} -i seq.fasta --outfmt=clu --resno --wrap=9000 --force'
p = subprocess.Popen(cmd.split(), stdout=subprocess.PIPE, stderr=subprocess.PIPE)
out = p.communicate()
os.remove('seq.fasta')
aln_data = out[0].decode('utf-8').split()
ref_aln = aln_data[6]
target_aln = aln_data[9]
counter_a = 0
counter_b = 0
numbering_dic = {}
for i in range(len(ref_aln)):
ref_char = ref_aln[i]
target_char = target_aln[i]
ref_resnum = list(reference_seq_dic[ref_chain])[counter_a]
try:
target_resnum = list(target_seq_dic[target_chain])[counter_b]
except IndexError:
# Target sequence exhausted, ignore
target_resnum = '-'
# print(ref_char, ref_resnum, target_char, target_resnum)
if '-' not in ref_char:
counter_a += 1
if '-' not in target_char:
counter_b += 1
if '-' not in ref_char and '-' not in target_char:
numbering_dic[target_resnum] = ref_resnum
identity = out[0].decode('utf-8').count('*') / float(len(ref_seq))
coverage = len(numbering_dic) / len(ref_aln)
# print(ref_chain, target_chain, identity, coverage)
# print(f'>R:{ref_chain}\n{ref_aln}')
# print(f'>T:{target_chain}\n{target_aln}')
try:
identity_dic[ref_chain].append((target_chain, identity, coverage, numbering_dic))
except KeyError:
identity_dic[ref_chain] = [(target_chain, identity, coverage, numbering_dic)]
# print('#########\n')
# do the renumbering
for i, ref_c in enumerate(reference_chains):
target_info_list = [(v[0], v[1], v[2]) for v in identity_dic[ref_c]]
# sort by identity and coverage
sorted_target_list = sorted(target_info_list, key=lambda x: (-x[2], x[1]))
# create a catalog with possible numbering references
numbering_dic_catalog = dict([(v[0], v[3]) for v in identity_dic[ref_c]])
if len(set([ e[1] for e in sorted_target_list])) == 1:
# this is a h**o-something, match is sequentialy
selected_chain = target_chains[i]
else:
# get the highest identity/coverage
selected_chain = sorted_target_list[0][0]
# select the correct numbering dictionary
selected_numbering_dic = numbering_dic_catalog[selected_chain]
# just for readability:
old_chain = selected_chain
new_chain = ref_c
# replace the target chain (old) with the same observed in the reference (new)
chain_matched_pdb = PDB.replace_chain(pdb, old_chain, new_chain)
# renumber!
# Note, if residue is present in target and not in reference it will be DELETED, use with caution
renumbered_pdb = PDB.renumber(chain_matched_pdb, selected_numbering_dic, new_chain, overwrite=True)
return True
def dockq(ref, pdb_f, dockq_exec):
irms = .0
lrms = .0
fnat = .0
dockq_score = .0
capri = None
order = None
ref = PDB.fix_id(ref)
pdb_f = PDB.fix_id(pdb_f)
reference_chains = PDB.identify_chainseg(ref)
pdb_chains = PDB.identify_chainseg(pdb_f)
result_dic = {}
if reference_chains != pdb_chains:
print(
f'+ WARNING: Skipping {pdb_f}, number of chains do not match. Expected {len(reference_chains)} found {len(pdb_chains)}'
)
interface_name = ''
result_dic[f'{interface_name}_irms'] = .0
result_dic[f'{interface_name}_lrms'] = .0
result_dic[f'{interface_name}_fnat'] = .0
result_dic[f'{interface_name}_capri'] = ''
result_dic[f'{interface_name}_dockq'] = .0
result_dic[f'{interface_name}_order'] = ''
else:
for comb in itertools.combinations(pdb_chains, 2):
# interface_name = ''.join(comb) + '-' + [e for e in pdb_chains if e not in comb][0]
interface_name = ''.join(comb)
cmd = f'{dockq_exec} {pdb_f} {ref} -native_chain1 {comb[0]} {comb[1]} -perm1'
p = subprocess.run(cmd.split(), stdout=subprocess.PIPE)
out = p.stdout.decode('utf-8').split('\n')
for l in out:
if l.startswith('Best score'):
output_l = l.split()
order = f'{output_l[-4]}->{output_l[-1]}'
elif l.startswith('Fnat'):
fnat = float(l.split()[1])
elif l.startswith('iRMS'):
irms = float(l.split()[1])
elif l.startswith('LRMS'):
lrms = float(l.split()[1])
elif l.startswith('DockQ_CAPRI'):
capri = l.split()[1]
elif l.startswith('DockQ'):
dockq_score = float(l.split()[1])
else:
pass
result_dic[f'{interface_name}_irms'] = irms
result_dic[f'{interface_name}_lrms'] = lrms
result_dic[f'{interface_name}_fnat'] = fnat
result_dic[f'{interface_name}_capri'] = capri
result_dic[f'{interface_name}_dockq'] = dockq_score
result_dic[f'{interface_name}_order'] = order
return result_dic
class TestPDB(unittest.TestCase):
def setUp(self):
self.PDB = PDB()
def test_treat_ensemble(self):
copyfile(f'{data_path}/mini_ens.pdb', f'{data_path}/temp_ens.pdb')
input_pdb_dic = {'mol1': f'{data_path}/temp_ens.pdb'}
treated_dic = self.PDB.treat_ensemble(input_pdb_dic)
expected_treated_dic = {'mol1': [f'{data_path}/temp_1.pdb', f'{data_path}/temp_2.pdb']}
self.assertEqual(treated_dic, expected_treated_dic)
self.assertTrue(filecmp.cmp(f'{data_path}/temp_1.pdb', f'{data_path}/mini_ens1.pdb'))
self.assertTrue(filecmp.cmp(f'{data_path}/temp_2.pdb', f'{data_path}/mini_ens2.pdb'))
os.remove(f'{data_path}/temp_1.pdb')
os.remove(f'{data_path}/temp_2.pdb')
os.remove(f'{data_path}/temp_ens.pdb')
def test_load_structure(self):
pdb_f = f'{data_path}/miniA.pdb'
pdb_dic = self.PDB.load_structure(pdb_f)
expected_pdb_dic = {'A': ['ATOM 2 CA MET A 1 16.967 12.784 4.338 1.00 10.80 A C \n',
'ATOM 9 CA ARG A 2 13.856 11.469 6.066 1.00 8.31 A C \n',
'ATOM 16 CA CYS A 3 13.660 10.707 9.787 1.00 5.39 A C \n']}
self.assertEqual(pdb_dic, expected_pdb_dic)
def test_identify_chains(self):
pdb_f = f'{data_path}/mini.pdb'
chain_l = self.PDB.identify_chains(pdb_f)
expected_chain_l = ['A', 'B','C']
self.assertEqual(chain_l, expected_chain_l)
def test_identify_segids(self):
pdb_f = f'{data_path}/miniA.pdb'
segid_l = self.PDB.identify_segids(pdb_f)
expected_segid_l = ['A']
self.assertEqual(segid_l, expected_segid_l)
def test_split_models(self):
ensamble_f = f'{data_path}/mini_ens.pdb'
model_list = self.PDB.split_models(ensamble_f)
expected_list = [f'{data_path}/mini_1.pdb', f'{data_path}/mini_2.pdb']
self.assertEqual(model_list, expected_list, 'Name of list elements differ')
self.assertTrue(filecmp.cmp(f'{data_path}/mini_1.pdb', f'{data_path}/mini_1.gold'))
self.assertTrue(filecmp.cmp(f'{data_path}/mini_2.pdb', f'{data_path}/mini_2.gold'))
for f in model_list:
os.remove(f)
def test_fix_id(self):
nosegid_pdb_f = f'{data_path}/mini.pdb'
nochain_pdb_f = f'{data_path}/mini_nochain.pdb'
segid_pdb = self.PDB.fix_id(nosegid_pdb_f, priority='chain', overwrite=False)
chain_pdb = self.PDB.fix_id(nochain_pdb_f, priority='seg', overwrite=False)
self.assertTrue(filecmp.cmp(segid_pdb, f'{data_path}/mini_segid.pdb'))
self.assertTrue(filecmp.cmp(chain_pdb, f'{data_path}/mini_segid.pdb'))
os.remove(f'{data_path}/mini.pdb_')
os.remove(f'{data_path}/mini_nochain.pdb_')
def test_add_chainseg(self):
copyfile(f'{data_path}/mini.pdb', f'{data_path}/temp.pdb')
check = self.PDB.add_chainseg(f'{data_path}/temp.pdb', 'A')
self.assertTrue(check)
self.assertTrue(filecmp.cmp(f'{data_path}/temp.pdb', f'{data_path}/miniA.pdb'))
os.remove(f'{data_path}/temp.pdb')
def test_identify_chainseg(self):
pdbf = f'{data_path}/miniA.pdb'
chainseg = self.PDB.identify_chainseg(pdbf)
self.assertEqual(chainseg, ['A'])
def test_fix_chainseg(self):
copyfile(f'{data_path}/mini_1.gold', f'{data_path}/mol1.pdb')
copyfile(f'{data_path}/mini_2.gold', f'{data_path}/mol2.pdb')
input_pdb_dic = {'mol1': f'{data_path}/mol1.pdb', 'segid1': 'X', 'mol2': f'{data_path}/mol2.pdb'}
return_pdb_dic = self.PDB.fix_chainseg(input_pdb_dic)
expected_return_dic = {'mol1': f'{data_path}/mol1.pdb', 'mol2': f'{data_path}/mol2.pdb'}
self.assertEqual(return_pdb_dic, expected_return_dic)
self.assertTrue(filecmp.cmp(f'{data_path}/mol1.pdb', f'{data_path}/miniX.pdb'))
os.remove(f'{data_path}/mol1.pdb')
os.remove(f'{data_path}/mol2.pdb')
def test_sanitize(self):
copyfile(f'{data_path}/mini.dirty.pdb', f'{data_path}/temp.pdb')
input_pdb_dic = {'mol1': [f'{data_path}/temp.pdb']}
model_list = self.PDB.sanitize(input_pdb_dic)
expected_model_list = [f'{data_path}/temp.pdb']
self.assertEqual(model_list, expected_model_list)
self.assertTrue(filecmp.cmp(f'{data_path}/temp.pdb', f'{data_path}/mini.clean.pdb'))
os.remove(f'{data_path}/temp.pdb')
def test_count_atoms(self):
pdb_f = f'{data_path}/mini.pdb'
atom_count = self.PDB.count_atoms(pdb_f)
self.assertEqual(atom_count, 3)
def test_organize_chains(self):
pass
def test_replace_chain(self):
pdb_f = f'{data_path}/mini.pdb'
newchain_pdb = self.PDB.replace_chain(pdb_f, 'A', 'X', overwrite=False)
self.assertTrue(filecmp.cmp(newchain_pdb, f'{data_path}/mini_A-X.pdb'))
os.remove(f'{data_path}/mini.pdb_')
def test_renumber(self):
pass
def test_load_seq(self):
pass
def tearDown(self):
pass
def prepare_input(pdb_input, psf_input=None):
""" Write input of recipe """
# This section will be written for any recipe
# Even if some CNS variables are not used, it should not be an issue.
input_str = '\n! Input structure\n'
string = 'ABCDEFGHIJKLMNOPQRSTUVWXYZ'
ncomp = None
if psf_input:
if type(psf_input) == str:
input_str += 'structure\n'
input_str += f' @@{psf_input}\n'
input_str += 'end\n'
if type(psf_input) == list:
input_str += 'structure\n'
for psf in psf_input:
input_str += f' @@{psf}\n'
input_str += 'end\n'
if type(pdb_input) == str:
ncomp = 1
if psf_input:
input_str += f'coor @@{pdb_input}\n'
else:
pass
# $file variable is still used by some CNS recipes, need refactoring!
input_str += f'eval ($file=\"{pdb_input}\")\n'
if type(pdb_input) == list or type(pdb_input) == tuple:
ncomp = len(pdb_input)
for pdb in pdb_input:
input_str += f'coor @@{pdb}\n'
chainsegs = PDB.identify_chainseg(pdb_input)
ncomponents = len(chainsegs)
input_str += f'eval ($ncomponents={ncomponents})\n'
for i, segid in enumerate(chainsegs):
input_str += f'eval ($prot_segid_mol{i+1}="{segid}")\n'
try:
ambig_fname = glob.glob('data/ambig.tbl')[0]
input_str += f'eval ($ambig_fname="{ambig_fname}")\n'
except IndexError:
input_str += f'eval ($ambig_fname="")\n'
try:
unambig_fname = glob.glob('data/unambig.tbl')[0]
input_str += f'eval ($unambig_fname="{unambig_fname}")\n'
except IndexError:
input_str += f'eval ($unambig_fname="")\n'
try:
hbond_fname = glob.glob('data/hbond.tbl')[0]
input_str += f'eval ($hbond_fname="{hbond_fname}")\n'
except IndexError:
input_str += f'eval ($hbond_fname="")\n'
try:
dihe_fname = glob.glob('data/dihe.tbl')[0]
input_str += f'eval ($dihe_fname="{dihe_fname}")\n'
except IndexError:
input_str += f'eval ($dihe_fname="")\n'
try:
tensor_fname = glob.glob('data/tensor.tbl')[0]
input_str += f'eval ($tensor_tbl="{tensor_fname}")\n'
except IndexError:
input_str += f'eval ($dihe_fname="")\n'
seed = random.randint(100, 999)
input_str += f'eval ($seed={seed})\n'
return input_str