def cyclesfit(atoms1, atoms2,iterations=5,cutoff=2):
"""
Performs a PyMol-style fit with rejection cycles
After every cycle, atoms with more than <cutoff> standard deviations error (in A2)
get rejected from the selection
"""
a1, a2 = numpy.array(atoms1), numpy.array(atoms2)
pivot = numpy.sum(a2,axis=0) / float(len(a2))
for n in range(iterations):
rotmat, offset, rmsd = rmsdlib.fit(a1,a2)
aa2 = apply_matrix(a2, pivot, rotmat, offset)
dif = aa2 - a1
d = dif.sum(axis=1)
sd = d * d
msd = sd.sum()/len(sd)
#rmsd = sqrt(msd)
std = numpy.std(sd)
if (std < 0.1): std = 0.1
keep = numpy.less(sd,cutoff*std)
aa1 = a1[keep]
aa2 = a2[keep]
a1 = aa1
a2 = aa2
return rmsdlib.fit(a1, a2)
def run(runarg):
mobile, outputfile = runarg
lines2, atoms2, extralines2 = read_pdb(mobile)
#select backbone
if args.allatoms:
assert not args.ca
atoms2_fit = atoms2
elif args.ca:
atoms2_fit = select_ca(lines2, atoms2)
else:
atoms2_fit = select_bb(lines2, atoms2)
assert len(atoms1_fit) and len(atoms1_fit) == len(atoms2_fit), (
len(atoms1_fit), len(atoms2_fit))
if args.iterative:
#perform a Pymol-style iterative fit
rotmat, offset, rmsd = cyclesfit(atoms1_fit, atoms2_fit,
args.iterative_cycles,
args.iterative_cutoff)
else:
#perform a direct fit
rotmat, offset, rmsd = rmsdlib.fit(atoms1_fit, atoms2_fit)
if args.rmsd:
return rmsd
else:
pivot = numpy.sum(atoms2_fit, axis=0) / float(len(atoms2_fit))
fitted_atoms = apply_matrix(atoms2, pivot, rotmat, offset)
write_pdb(outputfile, lines2, fitted_atoms, extralines2)
return outputfile
def run(runarg):
ref = runarg
print(ref)
lines_ref, atoms_ref, extralines_ref = read_pdb(mobiles[ref])
results = np.zeros(len(mobiles))
for ii in range(len(mobiles)):
lines, atoms, extralines = read_pdb(mobiles[ii])
rotmap, offset, rmsd = rmsdlib.fit(atoms_ref, atoms)
results[ii] = round(rmsd, 3)
return ref, results
def run2(runarg):
ref = runarg
print(ref)
atoms_ref = test[ref]
results = np.zeros(len(mobiles))
for ii in range(len(mobiles)):
if ii == ref:
rmsd = 0
else:
atoms = test[ii]
rotmap, offset, rmsd = rmsdlib.fit(atoms_ref, atoms)
results[ii] = round(rmsd, 3)
return ref, results
def apply_rnalib(pdb, lib, heavy):
"""
Adds missing atoms using an RNA lib
"""
assert heavy == True #TODO: remove this when the rna library has hydrogens
import numpy
syspath = list(sys.path)
sys.path.insert(0, os.environ["ATTRACTTOOLS"])
import rmsdlib
from scipy.spatial.distance import cdist
sys.path[:] = syspath
for res in pdb:
if res.resname not in ("RA", "RC", "RG", "RU"): continue
while 1: #keep fixing as long as we can
missing = set()
top = res.topology
for a in top.atomorder:
aa = a.upper()
if heavy and aa[0].startswith("H"): continue
if aa not in res.coords:
missing.add(aa)
if not missing: break
nuc = res.resname[1]
for fmode, fatoms in (
("base", lib.baseatoms[nuc]),
("sugar", lib.sugaratoms),
("nucleotide", lib.monoatoms[nuc]),
):
#we can fix if there are any missing atoms, and there are at least three non-lacking atoms
if any([(m in fatoms) for m in missing]) and \
len([a for a in fatoms if a in res.coords]) >= 3:
fixmode = fmode
break
else:
break #no more fixing to be done...
if fixmode == "base":
libcoor = lib.base[nuc][numpy.newaxis]
atoms = lib.baseatoms[nuc]
elif fixmode == "sugar":
libcoor = lib.sugar
atoms = lib.sugaratoms
elif fixmode == "nucleotide":
libcoor = lib.mono[nuc]
atoms = lib.monoatoms[nuc]
coor = numpy.array([res.coords[a] for a in atoms if a in res.coords])
mask = numpy.array([(a in res.coords) for a in atoms])
refecoor = numpy.compress(mask, libcoor, axis=1) #or: refecoor = libcoor[:,mask] (slower)
rotmat, offset, rmsd = rmsdlib.fit(refecoor[0],coor)
pivot = numpy.sum(coor,axis=0) / float(len(coor))
fitcoor = numpy.array(_apply_matrix(coor, pivot, rotmat, offset))
fitcoor = fitcoor.flatten()[numpy.newaxis]
refecoor = refecoor.reshape((len(refecoor), 3 * len(coor)))
d = cdist(fitcoor, refecoor, 'sqeuclidean')[0]
libconfnr = numpy.argmin(d)
libconf = libcoor[libconfnr]
libconf = _apply_matrix(libconf, pivot+offset, rotmat.T, -offset)
for anr, a in enumerate(atoms):
if a in missing or fixmode == "base":
x,y,z = libconf[anr]
res.coords[a] = x,y,z
if fixmode == "nucleotide": break
sys.stdout.flush()
if name is not None:
newargs = initargs + [
'--imodes', 'flexm-' + str(nstruc + 1) + name + '.dat'
]
if not os.path.exists('flexm-' + str(nstruc + 1) + name + '.dat'):
break
collectlib.collect_iattract(newargs)
result = collectlib.collect_next()
if result: break
nstruc += 1
coor = collectlib.collect_all_coor()
coor = numpy.compress(unbound_hmask, coor, axis=0)
fcoor = numpy.compress(selmask, coor, axis=0)
U = rmsdlib.fit(fcoor, fboundatoms)[0]
rotfcoor = numpy.array([], ndmin=2)
L = len(fcoor)
COM1 = numpy.sum(fcoor, axis=0) / float(L)
COM2 = numpy.sum(fboundatoms, axis=0) / float(L)
for c in fcoor:
rotfcoor = numpy.append(rotfcoor, U.dot(c - COM1))
rotfcoor = numpy.reshape(rotfcoor, (-1, 3))
f1.write(str(nstruc))
for start, end in ligandselmask:
sel1 = rotfcoor[start:end]
sel2 = fboundatoms[start:end]
sel2 = sel2 - COM2
irmsd = rmsdlib.rmsd(sel1, sel2)
f1.write(" %.3f" % irmsd)
imask = rmsdlib.build_interfacemask(bounds, thresh)
selmask = (selmask & imask)
if not opt_allatoms:
amask = rmsdlib.build_atommask(bounds, atomnames)
selmask = (selmask & amask)
fboundatoms = allboundatoms[selmask]
nstruc = 0
f1 = sys.stdout
if output is not None:
f1 = open(output, 'w')
while 1:
sys.stdout.flush()
if name is not None:
newargs = initargs + [
'--imodes', 'flexm-' + str(nstruc + 1) + name + '.dat'
]
if not os.path.exists('flexm-' + str(nstruc + 1) + name + '.dat'):
break
collectlib.collect_iattract(newargs)
result = collectlib.collect_next()
if result: break
nstruc += 1
coor = collectlib.collect_all_coor()
coor = numpy.compress(unbound_hmask, coor, axis=0)
fcoor = numpy.compress(selmask, coor, axis=0)
irmsd = rmsdlib.fit(fboundatoms, fcoor)[2]
f1.write(str(nstruc) + " %.3f\n" % irmsd)
coor = []
for p in unbounds:
p.remove_hydrogens()
for c in p.coordinates():
coor.append(c)
coor = numpy.array(coor)
if len(coor) == 0:
nstruc += 1
f1.write(str(nstruc)+" None\n")
continue
rmsdlib.check_pdbs(unbounds, bounds)
nstruc += 1
fcoor = numpy.compress(selmask, coor, axis=0)
U = rmsdlib.fit(fboundatoms,fcoor)[0]
rotfcoor = numpy.array([],ndmin=2)
for c in fcoor:
rotfcoor = numpy.append(rotfcoor,U.dot(c))
rotfcoor = numpy.reshape(rotfcoor,(-1,3))
f1.write(str(nstruc))
for start, end in ligandselmask:
sel1 = rotfcoor[start:end]
sel2 = fboundatoms[start:end]
irmsd = rmsdlib.rmsd(sel1,sel2)
f1.write(" %.3f" % irmsd)
f1.write("\n")
#select backbone if args.allatoms: assert not args.ca atoms1_fit = atoms1 atoms2_fit = atoms2 elif args.ca: atoms1_fit = select_ca(lines1, atoms1) atoms2_fit = select_ca(lines2, atoms2) else: atoms1_fit = select_bb(lines1, atoms1) atoms2_fit = select_bb(lines2, atoms2) assert len(atoms1_fit) and len(atoms1_fit) == len(atoms2_fit), (len(atoms1_fit), len(atoms2_fit)) if args.iterative: #perform a Pymol-style iterative fit rotmat, offset, rmsd = cyclesfit(atoms1_fit,atoms2_fit, args.iterative_cycles, args.iterative_cutoff) else: #perform a direct fit rotmat, offset, rmsd = rmsdlib.fit(atoms1_fit,atoms2_fit) pivot = numpy.sum(atoms2_fit,axis=0) / float(len(atoms2_fit)) fitted_atoms = apply_matrix(atoms2, pivot, rotmat, offset) if args.rmsd: print "%.3f" % rmsd else: write_pdb(lines2, fitted_atoms, extralines2)
coor = []
for p in unbounds:
p.remove_hydrogens()
for c in p.coordinates():
coor.append(c)
coor = numpy.array(coor)
if len(coor) == 0:
nstruc += 1
f1.write(str(nstruc) + " None\n")
continue
rmsdlib.check_pdbs(unbounds, bounds)
nstruc += 1
fcoor = numpy.compress(selmask, coor, axis=0)
U, t, irmsd = rmsdlib.fit(fcoor, fboundatoms)
rotfcoor = numpy.array([], ndmin=2)
L = len(fcoor)
COM1 = numpy.sum(fcoor, axis=0) / float(L)
COM2 = numpy.sum(fboundatoms, axis=0) / float(L)
for c in fcoor:
rotfcoor = numpy.append(rotfcoor, U.dot(c - COM1))
rotfcoor = numpy.reshape(rotfcoor, (-1, 3))
f1.write(str(nstruc))
for start, end in ligandselmask:
sel1 = rotfcoor[start:end, :]
sel2 = fboundatoms[start:end, :]
sel2 = sel2 - COM2
irmsd = rmsdlib.rmsd(sel1, sel2)
f1.write(" %.3f" % irmsd)