def append(self, newms):
""" add new molecules for training """
a3 = []
for mi in newms:
om = crc.RDMol(mi, forcefield=self.ff)
om.optg()
om.optg2()
a3.append(om.ats)
self.ms = self.a1 + self.a2 + a3 + self.ts
def diagnose1(fa, fq):
""" do diagnose for one amon and the query """
oa = crk.RDMol(fa)
oa.get_angles(wH=F, key='ia')
oa.get_dihedral_angles(wH=F, key='ia')
oq = crk.RDMol(fq)
oq.get_angles(wH=F, key='ia')
oq.get_dihedral_angles(wH=F, key='ia')
smi = oa.prop['smiles_indigo']
patt = crk.smi2patt(smi)
print('patt=', patt)
mp = Chem.MolFromSmarts(patt)
iass_a = oa.m.GetSubstructMatches(mp)
iass_q = oq.m.GetSubstructMatches(mp)
types_a, vals_a = get_angles34(oa.dangs, iass_a)
#dic_a = dict(zip(types_a[0],vals_a[0]))
types_q, _vals_q = get_angles34(oq.dangs, iass_q)
n = len(types_q)
vals_q = []
for i in range(n):
types = types_q[i]
vals = _vals_q[i]
#print ' -- types = ', types
dic_q = dict(list(zip(types, vals)))
_vals = []
for key in types_a[0]:
_vals.append(dic_q[key])
vals_q.append(_vals)
print(' amon:')
print(types_a[0])
print(list(vals_a[0]))
print(' query:')
for i in range(n):
print(vals_q[i])
print(' Difference:')
for i in range(n):
print(list(vals_q[i] - vals_a[0]))
def __init__(self, fs, rsmi, props=['HF'], istart=0):
self.nc0 = len(fs)
fsc = []
cso = [] # mol objects
ms = [] # ase mols
ys = []
#assert '_c' in fs[0]
#self.filename = '_'.join( fs[0].split('/')[-1].split('_')[:-1] )
self.fs_diss = []
self.fs_redundant = []
cids = []
for i, f in enumerate(fs):
fmt = f[-3:]
if fmt in ['log', 'out']: #G09 output file
dic = GR0(f, istart=istart)[-1]
zs = np.array(dic['Atomic_numbers'], np.int)
coords = np.array(dic['Positions'])
m = atoms(zs, coords)
try:
co = cmm.Mol(zs, coords, ican=True)
can2 = get_alternative(co.can)
if rsmi not in [co.can, can2]:
print("#ERROR: %s has a SMILES %s, differs from %s" %
(f, co.can, rsmi))
self.fs_diss.append(f)
continue
else:
_ys = {}
for key in props:
_ys[key] = get_val(dic, key)
ys.append(_ys)
ms.append(m)
cso.append(co)
fsc.append(f)
except:
print("#ERROR: this is a radical!")
self.fs_diss.append(f)
continue
elif fmt in ['mol', 'sdf']:
oo = crk.RDMol(f)
m = atoms(oo.zs, oo.coords)
ms.append(m)
cso.append(oo.prop['smiles_indigo'])
fsc.append(f)
ys.append([oo.prop[k] for k in props])
self.cso = cso
self.ms = ms
self.fsc = fsc
self.nc = len(cso)
self.ys = ys
def query(self, fq, idQ=None, k=7):
if self.cmaps[0]:
#assert len(fsq) > 0, '#ERROR: `fsq not specified!'
assert idQ is not None
self.idQ = idQ #fsq.index(fq)
self.fq = fq
#zs, coords, ydic = read_xyz_simple(f,opt='z')
objq = crk.RDMol(fq)
zs, coords, ydic = objq.zs, objq.coords, objq.prop
smi = ydic['smiles_%s' % self.ctk]
_ys = []
for key in self.props:
_yi = ydic[key] if key in ydic.keys() else np.nan
_ys.append(_yi)
ys_q = np.array([_ys]) #[ [ydic[key] for key in self.props ] ])
nheav = (np.array(zs) > 1).sum()
#ao = cma.amon(smi, k) # amon object. Note that can's are of indigo format
#assert ao.iok
#amons_q, ns_q, ats_q = ao.get_amons(iao=T) # idxs of atoms (in query mol) as output as well
reduce_namons = T
ao = coa.ParentMols([smi], reduce_namons, wg=F, imap=T, k=7)
#amons_q = []
#for cci in ao.cans:
# mobj = indigo.Indigo()
# m2 = mobj.loadMolecule(cci)
# amons_q.append( m2.canonicalSmiles() )
#self.ats_q = ao.atsa
self.amons_q = ao.cans #amons_q ## now stick to can of oechem format
#self.ns_q = ao.nsa
self.ys_q = ys_q
self.zs_q = np.array(zs, np.int)
self.nsheav_q = [nheav]
self.coords_q = np.array(coords)
# tor
#self.x_q = get_torsions( atoms(zs,coords) ) #; sys.exit()
objq.get_atypes()
objq.get_angles(wH=F, key='ia')
objq.get_dihedral_angles(wH=F, key='ia')
self.objq = objq
def __init__(self, obj, addh=True):
"""
process a list of SMILES objects
"""
smis = []
typ = type(obj)
if typ is str:
if os.path.exists(obj):
smis += [si.strip() for si in file(obj).readlines()]
else:
smis += [
obj,
]
elif typ is list:
for obj_i in obj:
if os.path.exists(obj_i):
smis += [si.strip() for si in file(obj).readlines()]
else:
smis += [
obj_i,
]
self.nm = len(smis)
self.objs = [crk.RDMol(si) for si in smis]
def __init__(self, objs, wd='targets/', reduce_namons=T, wg=F, \
i3d=T, a1=T, level=2, exta=0, k=7, k2=7, \
ff='mmff94', owt=F):
"""
vars
==============
level: extended amons level, 1 or 2
exta: maximal N_I of extended amons
owt: overwrite target (when writing sdf file for target)
"""
if isinstance(objs, str): objs = [objs]
if not os.path.exists(wd): os.mkdir(wd)
self.ff = ff
ts = []
fts = []
# first get 3d geom of targests
for obj in objs:
om = crc.RDMol(obj, forcefield=ff)
om.optg()
om.optg2()
ats = om.ats
can = coo.oem2can(coo.smi2oem(obj)[1])
can_hex = str2hex(coo.oem2can(coo.smi2oem(obj)[1]))
f1 = wd + '%s.xyz' % can #_hex
f2 = wd + '%s.sdf' % can #_hex
if owt or (not (os.path.exists(f2) and os.path.exists(f1))):
om.write_sdf(f2)
om.write_xyz(f1)
else:
ats = cc.obj2mol(f1, ['E'])
fts.append(f2)
ts.append(ats)
self.ts = ts
a1 = [] # generic amons
if a1:
imap = F if len(fts) == 1 else T
oa = coa.ParentMols(fts, reduce_namons, label=None, \
imap=imap, fixGeom=F, i3d=i3d, wg=wg, \
k=k,k2=k2, iprt=T, forcefield=ff, \
thresh=0.01, debug=F)
for mi in oa.ms:
om = crc.RDMol(mi, forcefield=ff)
#om.optg()
om.iFFOpt = T
om.optg2()
a.append(om.ats)
self.a1 = a1
a2 = []
assert exta >= 0, '#ERROR: `exta: N_I of extended amons shoud be >= 0'
if exta:
# add extended amons
oa2 = coae.transform(objs[0])
oa2.get_newolds()
newms, newms2 = oa2.T(level=level)
oa2.get_amons_extended(k=exta)
for mi in oa2.amons_extended:
om = crc.RDMol(mi, forcefield=ff)
om.optg()
om.optg2()
a2.append(om.ats)
self.a2 = a2
self.ms = a1 + a2 + ts
"OC12C3C4C1N4C32"] # the last one is highly strained, may have problem in acquring g0
elif n == 1:
f = args[0]
if f[-3:] in ['smi','can']:
objs = [ si.strip() for si in file(f).readlines() ]
else: # either an xyz file or a SMILES string
objs = args
else:
objs = args
isf = False
nobj = len(objs)
for i,obj in enumerate(objs):
if not os.path.isfile(obj):
f = tpf.NamedTemporaryFile(dir='/tmp').name + '.xyz'
m0 = cir.RDMol(obj, doff=True)
m0.write_xyz(f)
else:
isf = True
f = obj
o = cc.molecule(f, isimple=T)
can = 'None'
iok = T
if nobj > 1:
if trial:
try:
m = Mol(o.zs, o.coords, ican=True)
can = m.can
except:
iok = F #print(' conversion failed!')#pass
else:
def __init__(self, fd, fd2=None, fcanr=None, h5f=None, \
imb=False, props=['E'], ctk='oechem'):
"""
fcanr: file containing cononical SMILES for reference (from which conformers
were generated)
"""
self.ctk = ctk # aqml.cheminfomatic toolkit to be used, rdkit, indigo, openbabael or oechem
if fd[-1] == '/': fd = fd[:-1]
self.fda = fd
fs = io2.cmdout('ls %s/frag*.sdf' % fd)
self.fs = fs
nf = len(fs)
ns = [len(f) for f in fs]
assert len(np.unique(ns)) == 1, '#ERROR: filename length may differ'
assert nf > 0, '#ERROR: no *.sdf found'
objs = []
s = []
s1 = [] # unique smiles strings
_ncs = [0] * nf
imc = 0
for f in fs:
obj = crk.RDMol(f)
objs.append(obj)
smi = obj.prop['smiles_%s' % ctk]
s.append(smi)
if smi not in s1:
_ncs[imc] = 1
imc += 1
s1.append(smi)
else:
im = s1.index(smi)
_ncs[im] += 1
if fcanr is not None:
_amons = [si.strip() for si in open(fcanr).readlines()]
if _namon == len(s1):
print(
"#ERROR: #amons from sdf's inconsistent with that in %s" %
fcanr)
print(
" This means that some amons (graph) may have been")
print(" hiscarded due to dissociation")
raise
else:
_amons = s1
_namon = len(_amons)
self.amons = _amons
nm = len(s1)
ncs = _ncs[:nm]
cidxs = np.arange(nf).astype(int)
c2amap = []
ics2 = np.array(np.cumsum(ncs), dtype=int)
ics1 = np.array(np.concatenate(([0], ics2[:-1])), dtype=int)
for im in range(nm):
c2amap += [
im,
] * ncs[im]
a2cmap = [list(cidxs[ics1[iamon]:ics2[iamon]]) for iamon in range(nm)]
cmaps = [F]
# if provided h5 file, read conformer IDs from it
if h5f is not None:
dt = dd.io.load(h5f)
_maps = dt['maps']
if np.max(_maps) != nf - 1:
print(' * Error: #sdf files .ne. max(maps)+1!!')
raise
cmaps = [T, _maps]
# add new conformers from another folder
ic2 = nf
if fd2 is None:
c2amap = np.array(c2amap, dtype=int)
else:
fs2 = io2.cmdout('ls %s/frag*sdf' % fd2)
fs += fs2
nf2 = len(fs2)
s2 = []
ic2s = []
im2s = []
for f2 in fs2:
obj2 = crk.RDMol(f2)
smi = obj2.prop['smiles_%s' % self.ctk]
s2.append(smi)
if smi in _amons:
im2 = _amons.index(smi)
c2amap.append(im2)
a2cmap[im2] += [ic2]
ic2 += 1
objs.append(obj2)
assert len(s2) == nf2
c2amap = np.array(c2amap, dtype=int)
if cmaps[0]:
# now update maps file
_maps = cmaps[1]
nt, ncmax = _maps.shape
icsmax = np.arange(ncmax).astype(int)
_maps2 = []
ncs_added = []
for j in range(nt):
fil = (_maps[j] > -1)
ims_j = np.unique(c2amap[_maps[j][fil]])
ics_j = []
for k in ims_j:
#print('k,len(a2cmap)=',k,len(a2cmap))
t2 = np.array(a2cmap[k], dtype=int)
ics_j += list(t2[t2 >= nf])
ncs_added.append(len(ics_j))
_maps2.append(list(_maps[j][fil]) + ics_j)
nadd = max(ncs_added)
print(
' ## for some query molecule, %d conformers at most are added as new amons'
% nadd)
maps2 = -1 * np.ones((nt, ncmax + nadd), dtype=int)
for j in range(nt):
ncj = len(_maps2[j])
maps2[j, :ncj] = _maps2[j]
cmaps = [T, maps2.astype(int)]
c2amap = np.array(c2amap, dtype=int)
self.c2amap = c2amap # (amon/conformer) to (molecule graph) map
self.a2cmap = a2cmap
self.cmaps = cmaps
# if `fcanr is not None, check dissociated molecules
#if fcanr is not None:
# _remap = np.arange(nf).astype(int)
# for i,f in enumerate(fs):
# im = int(f.split('frag_')[1][:-4].split('_')[0])-1 # molecule index (graph! not conformer idx)
# smi = s[i]
# if smi != _amons[im]: # graph may have changed after optg
# _remap[i] = -1
nct = ic2
self.nmt = nm
self.nct = nct
self.fs = fs
objsc = []
ys = []
zs = []
nsheav = []
nas = []
coords = []
for i in range(nct):
objc = objs[i]
zs += list(objc.zs)
nas.append(objc.na)
coords += list(objc.coords)
nsheav.append(objc.nheav)
if imb: # calc many-body terms
objc.get_angles(wH=F, key='ia')
objc.get_dihedral_angles(wH=F, key='ia')
objsc.append(objc)
#print('f,e_qmc=',fs[i],objc.prop['QMC_B3LYP'])
ys.append([objc.prop[key] for key in props])
self.nas = np.array(nas, np.int)
self.zs = np.array(zs, np.int)
self.coords = np.array(coords)
self.nsheav = np.array(nsheav, np.int)
self.objsc = objsc
self.ics1 = ics1
self.ics2 = ics2
self.props = props
self.ys = np.array(ys)
