def copyDataFromTarget(target, model, labels=('psipred', 'netsurfp')):
data = {}
for label in target.getDataLabels():
if not label.startswith(tuple(labels)):
continue
data[label] = np.zeros(model.numAtoms(), dtype=target.getData(label).dtype)
for tchain in target.iterChains():
#try to find matching chain first
if tchain.getChid() in [x.getChid() for x in model.iterChains()]:
mchain = tchain.getChid()
else:
mchain = prody.matchChains(model, target[tchain.getChid()],
seqid=50, overlap=20)[0][0]
mchain = mchain.copy().getChids()[0]
for tres in tchain.ca.copy().iterResidues():
for label in data:
try:
indices = model[mchain][tres.getResnum()].getIndices()
data[label][indices] = tres.ca.getData(label)
except AttributeError: #target may have more residues than
#the model
pass
for label in data:
if data[label].dtype.char == 'S':
data[label][data[label] == ''] = '-'
model.setData(label, data[label])
def fix_openmm():
# get the whole crystal structure
# get only the ATOM records
# and HETAM records for MSE
# convert MSE to MET
with open('no_smet.pdb', 'w') as outfile:
with open('experimental.pdb') as infile:
for line in infile:
if line.startswith('ATOM'):
outfile.write(line)
if line.startswith('HETATM'):
if line[17:20] == 'MSE':
atom_name = line[12:17]
if atom_name == 'SE ':
atom_name = ' SD '
line_fixed = 'ATOM ' + line[
6:12] + atom_name + 'MET' + line[20:67] + '\n'
outfile.write(line_fixed)
# load the file into prody
p = prody.parsePDB('no_smet.pdb')
p = p.select('not hydrogen')
# get one of the rosetta models
r = prody.parsePDB('rosetta.pdb')
# perform an alignment to find out what part of the crystal structure
# corresponds to the rosetta file
match = prody.matchChains(r, p, subset='all', overlap=25,
pwalign=True)[0][1]
print len(match)
prody.writePDB('chain.pdb', match)
# now clean it up with pdb fixer
subprocess.check_call('python ~/Source/PdbFixer/pdbfixer.py chain.pdb',
shell=True)
# now load it with zam
p = protein.Protein('output.pdb')
p.Dehydrogen()
disulfide_pairs = find_disulfide(p)
for r1, r2 in disulfide_pairs:
print ' added disulfide between {} and {}'.format(r1, r2)
p.Res[r1].FullName = 'CYX'
p.Res[r2].FullName = 'CYX'
p.WritePdb('start.pdb')
# now run tleap
print ' running tleap'
run_tleap(disulfide_pairs)
def calc_pocket_rmsd(rec, lig, root):
"""
Calculate difference between the ligand reference receptor and
the receptor it is being docked into.
From original script by David Koes
"""
ligrec = lig.replace("LIG_aligned.sdf", "PRO.pdb")
rec = prody.parsePDB(os.path.join(root, rec))
ligrec = prody.parsePDB(os.path.join(root, ligrec))
lig = next(pybel.readfile("sdf", os.path.join(root, lig)))
c = np.array([a.coords for a in lig.atoms])
nearby = rec.select("protein and same residue as within 3.5 of point",
point=c)
matches = []
for cutoff in range(90, 0, -10):
# can't just set a low cutoff since we'll end up with bad alignments
# try a whole bunch of alignments to maximize the likelihood we get the right one
m = prody.matchChains(rec,
ligrec,
subset="all",
overlap=cutoff,
seqid=cutoff,
pwalign=True)
if m:
matches += m
minrmsd = np.inf
minbackrmsd = np.inf
for rmap, lrmap, _, _ in matches:
try:
closeatoms = set(nearby.getIndices())
lratoms = []
ratoms = []
for i, idx in enumerate(rmap.getIndices()):
if idx in closeatoms:
lratoms.append(lrmap.getIndices()[i])
ratoms.append(idx)
if len(lratoms) == 0:
continue
rmsd = prody.calcRMSD(rec[ratoms], ligrec[lratoms])
backrmsd = prody.calcRMSD(rec[ratoms] & rec.ca,
ligrec[lratoms] & ligrec.ca)
if rmsd < minrmsd:
minrmsd = rmsd
minbackrmsd = backrmsd
except:
pass
return minrmsd, minbackrmsd
def fix_openmm():
# get the whole crystal structure
# get only the ATOM records
# and HETAM records for MSE
# convert MSE to MET
with open('no_smet.pdb', 'w') as outfile:
with open('experimental.pdb') as infile:
for line in infile:
if line.startswith('ATOM'):
outfile.write(line)
if line.startswith('HETATM'):
if line[17:20] == 'MSE':
atom_name = line[12:17]
if atom_name == 'SE ':
atom_name = ' SD '
line_fixed = 'ATOM ' + line[6:12] + atom_name + 'MET' + line[20:67] + '\n'
outfile.write(line_fixed)
# load the file into prody
p = prody.parsePDB('no_smet.pdb')
p = p.select('not hydrogen')
# get one of the rosetta models
r = prody.parsePDB('rosetta.pdb')
# perform an alignment to find out what part of the crystal structure
# corresponds to the rosetta file
match = prody.matchChains(r, p, subset='all', overlap=25, pwalign=True)[0][1]
print len(match)
prody.writePDB('chain.pdb', match)
# now clean it up with pdb fixer
subprocess.check_call('python ~/Source/PdbFixer/pdbfixer.py chain.pdb', shell=True)
# now load it with zam
p = protein.Protein('output.pdb')
p.Dehydrogen()
disulfide_pairs = find_disulfide(p)
for r1, r2 in disulfide_pairs:
print ' added disulfide between {} and {}'.format(r1, r2)
p.Res[r1].FullName = 'CYX'
p.Res[r2].FullName = 'CYX'
p.WritePdb('start.pdb')
# now run tleap
print ' running tleap'
run_tleap(disulfide_pairs)
def calcS2(model_list, S2_records, S2_type, fit, fit_range):
"""Returns a dictonary with the average S2 values:
S2_calced[residue] = value"""
# fitting models
reference = model_list[0]
if fit and not PDB_model.is_fitted:
print("Start FITTING")
for i in range(1, len(model_list)):
mobile = model_list[i]
matches = prody.matchChains(reference, mobile)
match = matches[0]
ref_chain = match[0]
mob_chain = match[1]
if fit_range:
weights = np.zeros((len(ref_chain), 1), dtype=np.int)
fit_start, fit_end = fit_range.split('-')
for i in range(int(fit_start) - 1, int(fit_end) - 1):
weights[i] = 1
else:
weights = np.ones((len(ref_chain), 1), dtype=np.int)
t = prody.calcTransformation(mob_chain, ref_chain, weights)
t.apply(mobile)
PDB_model.is_fitted = True
# get NH vectors from models (model_data[] -> vectors{resnum : vector})
model_data = []
s2_pairs = {'N': 'H', 'CA': 'HA'}
for model in model_list:
current_Resindex = 1
has_first, has_second = False, False
vectors = {}
for atom in model:
# why not .getResnum() ???
atom_res = atom.getResindex() + 1
if atom_res != current_Resindex:
current_Resindex = atom_res
has_first, has_second = False, False
if atom_res == current_Resindex:
if atom.getName() == S2_type:
has_second = True
N_coords = Vec_3D(atom.getCoords())
elif atom.getName() == s2_pairs[S2_type]:
has_first = True
H_coords = Vec_3D(atom.getCoords())
if has_first and has_second:
has_first, has_second = False, False
vectors[atom_res] = Vec_3D(N_coords - H_coords).normalize()
model_data.append(vectors)
S2_calced = {}
# iterating over STR records
for resnum in [int(s2rec.resnum) for s2rec in S2_records]:
x2, y2, z2, xy, xz, yz = 0, 0, 0, 0, 0, 0
# iterating over PDB models
for m in model_data:
# coordinates in model at a given resnum
x, y, z = m[resnum].v[0], m[resnum].v[1], m[resnum].v[2]
x2 += x**2
y2 += y**2
z2 += z**2
xy += x * y
xz += x * z
yz += y * z
x2 /= len(model_data)
y2 /= len(model_data)
z2 /= len(model_data)
xy /= len(model_data)
xz /= len(model_data)
yz /= len(model_data)
# S2 calcuation
s2 = 3 / 2.0 * (x2**2 + y2**2 + z2**2 + 2 * xy**2 + 2 * xz**2 +
2 * yz**2) - 0.5
S2_calced[resnum] = s2
return S2_calced
def s2_values(
model_data, calculate_on_models, s2_records, s2_type, fit, fit_range
):
"""Returns a dictionary with the average S2 values:
s2_calced[residue] = value"""
if fit:
reference = model_data.atomgroup[:]
model_data.atomgroup.setACSIndex(0)
prody.alignCoordsets(model_data.atomgroup.calpha)
if fit_range:
for model_num in calculate_on_models:
model_data.atomgroup.setACSIndex(model_num)
mobile = model_data.atomgroup[:]
matches = prody.matchChains(reference, mobile)
match = matches[0]
ref_chain = match[0]
mob_chain = match[1]
weights = np.zeros((len(ref_chain), 1), dtype=np.int)
fit_start, fit_end = fit_range.split("-")
for i in range(int(fit_start) - 1, int(fit_end) - 1):
weights[i] = 1
t = prody.calcTransformation(mob_chain, ref_chain, weights)
t.apply(mobile)
# get NH vectors from models (model_data[] -> vectors{resnum : vector})
vector_data = []
s2_pairs = {"N": "H", "CA": "HA"}
h_coords = None
n_coords = None
for model_num in calculate_on_models:
model_data.atomgroup.setACSIndex(model_num)
current_resindex = 1
has_first, has_second = False, False
vectors = {}
for atom in model_data.atomgroup:
atom_res = atom.getResnum()
if atom_res != current_resindex:
current_resindex = atom_res
has_first, has_second = False, False
if atom_res == current_resindex:
if atom.getName() == s2_type:
has_second = True
n_coords = Vec3D(atom.getCoords())
elif atom.getName() == s2_pairs[s2_type]:
has_first = True
h_coords = Vec3D(atom.getCoords())
if has_first and has_second:
has_first, has_second = False, False
vectors[atom_res] = Vec3D(
n_coords - h_coords
).normalize()
vector_data.append(vectors)
s2_calced = {}
# iterating over STR records
for resnum in [int(s2rec.resnum) for s2rec in s2_records]:
x2, y2, z2, xy, xz, yz = 0, 0, 0, 0, 0, 0
# iterating over PDB models
for m in vector_data:
# coordinates in model at a given resnum
x, y, z = m[resnum].v[0], m[resnum].v[1], m[resnum].v[2]
x2 += x ** 2
y2 += y ** 2
z2 += z ** 2
xy += x * y
xz += x * z
yz += y * z
x2 /= len(vector_data)
y2 /= len(vector_data)
z2 /= len(vector_data)
xy /= len(vector_data)
xz /= len(vector_data)
yz /= len(vector_data)
s2 = (
3
/ 2.0
* (
x2 ** 2
+ y2 ** 2
+ z2 ** 2
+ 2 * xy ** 2
+ 2 * xz ** 2
+ 2 * yz ** 2
)
- 0.5
)
s2_calced[resnum] = s2
return s2_calced
def align():
global wd
ans = wd + '/challengedata/answers'
if os.path.isdir(
ans) == False: #if the answers directory isnt formed make it
os.mkdir(wd + '/challengedata/answers')
rddir = wd + '/challengedata/rdkit-scripts'
if os.path.isdir(rddir) == False:
a = 'git clone https://github.com/dkoes/rdkit-scripts'
os.system(a)
data = os.listdir(wd + '/challengedata')
for x in (data): #for each weeks data
if x == "readme.txt" or x == "latest.txt" or x == "answers" or x == "rdkit-scripts" or x == 'PDBfiles' or x == 'visual.txt':
pass
else:
toDir = wd + '/challengedata/answers/' + x
if os.path.isdir(
toDir
) == False: #if the path to answers dir doesnt exist
os.mkdir(toDir) #make directory
dock = os.listdir(wd + '/challengedata/' + x)
for y in (dock):
a = str(os.getcwd() + '/answers/' + x + '/' + y +
'/lmcss_docked.sdf')
if y == 'readme.txt' or y == 'new_release_structure_sequence_canonical.tsv' or y == 'new_release_structure_nonpolymer.tsv' or y == 'new_release_crystallization_pH.tsv' or y == 'new_release_structure_sequence.tsv':
pass
elif (os.path.isfile(a) == True):
pass
else:
input = os.listdir(wd + '/challengedata/' + x + '/' +
y)
for z in (input):
if z.startswith("LMCSS") and z.endswith(".pdb"):
if (z.endswith("lig.pdb")):
pass
else:
id = z.strip('.pdb')
sts = str("grep ATOM " + z +
" > lmcss_rec.pdb"
) #creates receptor .pdb file
cd = wd + '/challengedata'
os.chdir(
cd + '/' + x + '/' +
y) #change directory to week/ligand
os.system(
sts
) #runs and creates receptor .pbd file
os.chdir(cd) #back to challenge directory
input = os.listdir(
cd + '/' + x + '/' + y
) #lists files inside ligand in certain week
for z in (input):
if z.endswith(
".smi"
): # changes .smi -> lig.sdf
cd = str(os.getcwd())
sts = str(" " + cd + '/' + x +
'/' + y + '/' + z +
" lig.sdf --maxconfs 1")
os.chdir(cd + '/' + x + '/' + y)
os.system(
cd +
'/rdkit-scripts/rdconf.py' +
sts)
os.chdir(cd)
for z in (input): # runs smina
if z.endswith("lig.pdb"):
sts = str(
"smina -r lmcss_rec.pdb -l lig.sdf --autobox_ligand "
+ z + " -o " + id +
"_docked.sdf")
cd = str(
os.getcwd()) #lignad directory
os.chdir(cd + '/' + x + '/' + y)
#os.system(sts)
sts = str(
"smina -r lmcss_rec.pdb -l lig.sdf --autobox_ligand "
+ z + " -o lmcss_docked.sdf")
cd = str(
os.getcwd()) #lignad directory
os.chdir(cd + '/' + x + '/' + y)
os.system(sts)
os.chdir(cd)
cur = str(os.getcwd() + '/answers/' + x +
'/' + y)
if (os.path.isdir(cur) == True):
os.chdir(cd + '/' + x + '/' + y)
os.getcwd() ##
input = os.listdir(cd + '/' + x + '/' +
y)
for i in (input):
if i.endswith(
".txt"
) and i != "center.txt" and i != "visual.txt":
f = open(i)
lines = f.readlines()
ligand = lines[2].strip(
'ligand, ')
ligand = ligand.replace(
'\n', '')
ligand = str(ligand)
#gets the ligand from txt file
if i.endswith("lig.pdb"):
#see if pdb exists
prody.fetchPDB(y)
proteinPDB = prody.parsePDB(y)
ourPDB = prody.parsePDB(
'lmcss_rec.pdb')
a, b, seqid, overlap = prody.matchChains(
proteinPDB, ourPDB)[0]
b, protein_sp = prody.superpose(
b, a, weights=None)
b.select(ligand +
'_ligand.pdb')
sts = str("obrms -f " + i +
' ' + id +
"_docked.sdf")
#run obrms
# parse results and output to the visualization txt file
os.system(sts)
f = open('visual.txt', 'ab+')
f.write(x + ' smina ' + y +
'\n')
f.close
curdir = str(cd + '/' + x +
'/' + y + '/' +
id +
'_docked.sdf')
print(input) ##
for i in (input):
if i.endswith("lig.pdb"):
#see if pdb exists
protein = prody.fetchPDB(y)
#NEED NUMPY ARRAY
prody.writeArray(
'lmcss_docked_array.sdf',
array)
prody.superpose(
'lmcss_docked.sdf',
protein,
weights=None)
sts = str("obrms -f " + i +
" lmcss_docked.sdf")
#run obrms
# parse results and output to the visualization txt file
os.system(sts)
os.chdir(wd +
'/challengedata/')
f = open('visual.txt', 'ab+')
f.write(x + ' smina ' + y +
'\n')
f.close
curdir = str(
cd + '/' + x + '/' + y +
'/lmcss_docked.sdf')
todir = str(cd + '/answers/' +
x + '/' + y + '/')
shutil.copy(curdir, todir)
print(curdir)
break
os.chdir(wd)
else:
os.mkdir(cur)
os.chdir(cd + '/' + x + '/' + y)
input = os.listdir(cd + '/' + x + '/' +
y)
for i in (input):
if i.endswith(
".txt"
) and i != "center.txt" and i != "visual.txt":
f = open(i)
lines = f.readlines()
ligand = lines[2].strip(
"ligand, ")
ligand = ligand.replace(
'\n', '')
ligand = str(ligand)
#gets ligand from txt file
if i.endswith("lig.pdb"):
prody.fetchPDB(y)
proteinPDB = prody.parsePDB(y)
ourPDB = prody.parsePDB(
'lmcss_rec.pdb')
prody.matchChains(
proteinPDB, ourPDB)
protein_sp = prody.superpose(
ourPDB,
proteinPDB,
weights=None)
protein_sp.select(
ligand + '_ligand.pdb')
sts = str("obrms -f " + i +
' ' + id +
"_docked.sdf")
os.system(sts)
f = open('visual.txt', 'ab+')
f.write(x + ' smina ' + y +
'\n')
f.close
curdir = str(cd + '/' + x +
'/' + y + '/' +
id +
'_docked.sdf')
if i.endswith("lig.pdb"):
protein = prody.fetchPDB(y)
prody.writeArray(
'lmcss_docked_array.sdf',
array)
prody.superpose(
'lmcss_docked.sdf',
protein,
weights=None)
sts = str("obrms -f " + i +
" lmcss_docked.sdf")
os.system(sts)
os.chdir(wd +
'/challengedata/')
f = open('visual.txt', 'ab+')
f.write(x + ' smina ' + y +
'\n')
f.close
curdir = str(
cd + '/' + x + '/' + y +
'/lmcss_docked.sdf')
todir = str(cd + '/answers/' +
x + '/' + y + '/')
shutil.copy(curdir, todir)
print(curdir)
break
os.chdir(wd)