def convert_extension(infile, outfile, canonical=False):
"""
Convert one molecule file format into another using OpenEye tools.
The user may also assign canonical smiles as name before writing output.
"""
# open input file
mols = reader.read_mols(infile)
# open output file
ofs = oechem.oemolostream()
if not ofs.open(outfile):
oechem.OEThrow.Fatal("Unable to open %s for writing" % outfile)
# write to output
for mol in mols:
if canonical:
smi = oechem.OEMolToSmiles(mol)
for conf in mol.GetConfs():
if canonical:
conf.SetTitle(smi)
oechem.OEWriteConstMolecule(ofs, conf)
# close filestreams
ofs.close()
def basic_plot(infile,
tag,
style,
molname=None,
take_relative=False,
har_to_kcal=False):
"""
TODO
Parameters
----------
infile : string
Name of SDF file with information in SD tags.
tag : string
Full tag string directly as listed in the SD file.
style : string
plot style. can be 'scatter', 'line', or 'bar'
TODO
take_relative : Boolean
subtract lowest value
har_to_kcal : Boolean
multiply data in Hartrees by 627.5095 to yield kcal/mol
"""
# Open molecule file.
mols = reader.read_mols(infile)
for i, mol_i in enumerate(mols):
if molname is not None and mol_i.GetTitle() != molname:
continue
# get array of all conformer data of this mol
try:
data_array = np.fromiter(pt.get_sd_list(mol_i,
datum='',
taglabel=tag),
dtype=np.float64)
except ValueError:
data_array = np.asarray([np.nan]) * mol_i.NumConfs()
# exclude conformers for which job did not finish (nan)
nanIndices = np.argwhere(np.isnan(data_array))
for j in reversed(nanIndices): # loop in reverse to delete correctly
data_array = np.delete(data_array, j)
if take_relative:
data_array = data_array - np.amin(data_array)
if har_to_kcal:
data_array = 627.5095 * data_array
# generate plot
plt.plot(data_array)
plt.grid()
plt.title(mol_i.GetTitle() + '\n' + tag, fontsize=14)
plt.savefig(f'output_{i}.png', bbox_inches='tight')
plt.show()
def filter_confs(infile, tag, outfile):
"""
Read in OEMols (and each of their conformers) in 'infile'.
For each molecule:
rough filter conformers based on energy differences specified by 'tag',
fine filter conformers based on RMSD values.
Parameters
----------
infile : str
Name of SDF file with conformers to be filtered
tag : str
SD tag name with the energy value to roughly screen conformers before RMSD
Screening works by removing conformers of very similar energies, where
"similar" is defined by thresE parameter. Examples:
- "QM Psi4 Final Opt. Energy (Har) mp2/def-sv(p)"
- "QM Psi4 Final Single Pt. Energy (Har) mp2/def-sv(p)"
outfile : str
Name of the output file with filtered conformers
"""
# Parameters for distinguishing cutoff of conformer similarity
thresE = 5.E-4 # declare confs diff & skip RMSD comparison above this threshold
thresRMSD = 0.2 # above this threshold (Angstrom), confs are "diff" minima
wdir, fname = os.path.split(infile)
numConfsF = open(os.path.join(os.getcwd(), "numConfs.txt"), 'a')
numConfsF.write("\n{}\n".format(tag))
# open molecule file
rmsd_molecules = reader.read_mols(infile)
# Open outstream file.
rmsd_ofs = oechem.oemolostream()
if os.path.exists(outfile):
raise FileExistsError("Output file {} already exists in {}".format(
outfile, os.getcwd()))
if not rmsd_ofs.open(outfile):
oechem.OEThrow.Fatal("Unable to open %s for writing" % outfile)
# Identify minima and write output file.
for mol in rmsd_molecules:
if identify_minima(mol, tag, thresE, thresRMSD):
numConfsF.write("%s\t%s\n" % (mol.GetTitle(), mol.NumConfs()))
oechem.OEWriteConstMolecule(rmsd_ofs, mol)
else:
numConfsF.write("%s\t0\n" % (mol.GetTitle()))
numConfsF.close()
rmsd_ofs.close()
print("Done filtering %s to %s.\n" % (fname, outfile))
def convert_extension_separate(infile,
presuffix,
canonical=False,
separate='mol'):
"""
Convert one molecule file format into another using OpenEye tools.
The user may also assign canonical smiles as name before writing output.
Separate output into (each mol with all confs) or (each conf).
presuffix : list
first item contains prefix of output name, last item contains extension
ex. ['alkyl', '.xyz']
separate : string
'mol' or 'conf'
"""
# open input file
mols = reader.read_mols(infile)
# write to output
for i, mol in enumerate(mols):
# open output file
if separate == 'mol':
ofs = oechem.oemolostream()
if not ofs.open('{}_{}{}'.format(presuffix[0], str(i),
presuffix[1])):
oechem.OEThrow.Fatal("Unable to open %s for writing" % outfile)
if canonical:
smi = oechem.OEMolToSmiles(mol)
for j, conf in enumerate(mol.GetConfs()):
# open output file
if separate == 'conf':
ofs = oechem.oemolostream()
if not ofs.open('{}_{}_{}{}'.format(presuffix[0], str(i),
str(j), presuffix[1])):
oechem.OEThrow.Fatal("Unable to open %s for writing" %
outfile)
if canonical:
conf.SetTitle(smi)
oechem.OEWriteConstMolecule(ofs, conf)
# close filestreams
ofs.close()
def confs_to_psi(insdf,
method,
basis,
calctype='opt',
memory=None,
via_json=False):
"""
Read in molecule(s) (and conformers, if present) in insdf file. Create
Psi4 input calculations for each structure.
Parameters
----------
insdf: string
Name of the molecule file for which to create Psi4 input file.
SDF format can contain multiple molecules and multiple conformers per
molecule in a single file.
method: string
Name of the method as understood by Psi4. Example: "mp2"
basis : string
Name of the basis set as understood by Psi4. Example: "def2-sv(p)"
calctype : string
What kind of Psi4 calculation to run. Supported inputs are:
'opt' for geometry optimization,
'spe' for single point energy calculation, and
'hess' for Hessian calculation.
memory : string
How much memory each Psi4 job should take. If not specified, the
default in Psi4 is 500 Mb. Examples: "2000 MB" "1.5 GB"
http://www.psicode.org/psi4manual/master/psithoninput.html
via_json : Boolean
If True, use JSON wrapper for Psi4 input and output.
- Psi4 input would be in "input.py", called with python
- Psi4 output would be in "output.json"
If False, use normal text files for Psi4 input and output.
- Psi4 input would be in "input.dat"
- Psi4 output would be in "output.dat"
"""
wdir = os.getcwd()
# open molecules
molecules = reader.read_mols(insdf)
### For each molecule: for each conf, generate input
for mol in molecules:
print(mol.GetTitle(), mol.NumConfs())
if not mol.GetTitle():
sys.exit("ERROR: OEMol must have title assigned! Exiting.")
for i, conf in enumerate(mol.GetConfs()):
# change into subdirectory ./mol/conf/
subdir = os.path.join(wdir, "%s/%s" % (mol.GetTitle(), i + 1))
if not os.path.isdir(subdir):
os.makedirs(subdir)
if os.path.exists(os.path.join(subdir, 'input.dat')):
print("Input file already exists. Skipping.\n{}\n".format(
os.path.join(subdir, 'input.dat')))
continue
label = mol.GetTitle() + '_' + str(i + 1)
if via_json:
ofile = open(os.path.join(subdir, 'input.py'), 'w')
ofile.write("# molecule {}\n\nimport numpy as np\nimport psi4"
"\nimport json\n\njson_data = ".format(label))
json.dump(make_psi_json(conf, label, method, basis, calctype,
memory),
ofile,
indent=4,
separators=(',', ': '))
ofile.write(
"\njson_ret = psi4.json_wrapper.run_json(json_data)\n\n")
ofile.write("with open(\"output.json\", \"w\") as ofile:\n\t"
"json.dump(json_ret, ofile, indent=2)\n\n")
else:
ofile = open(os.path.join(subdir, 'input.dat'), 'w')
ofile.write(
make_psi_input(conf, label, method, basis, calctype,
memory))
ofile.close()
def extract_enes(dict1, mol_slice=[]):
"""
From files in input dictionaries, read in molecules, extract information
from SD tags for conformer energies and indices.
Parameters
----------
dict1 : dict
dictionary of input files and information to extract from SD tags
keys are: 'theory' 'fname' 'tagkey' 'label'
mol_slice : list
list of indices from which to slice mols generator for read_mols
[start, stop, step]
Returns
-------
titleMols : list of strings
names of all molecules in the SDF file
confNums : list of ints
conformer index numbers
enes : list of numpy arrays
conformer energies of the compared file (kcal/mol)
confNans : list of numpy arrays
indices of enes where the values are nans
"""
# Open molecule file.
if len(mol_slice) == 3:
mols = reader.read_mols(dict1['fname'], mol_slice)
else:
mols = reader.read_mols(dict1['fname'])
short_tag = dict1['tagkey']
qmethod, qbasis = reader.separated_theory(dict1['theory'])
titleMols = []
confNums = []
enes = []
confNans = []
for imol in mols:
# Get absolute energies from the SD tags
iabs = np.array(
list(
map(float,
pt.get_sd_list(imol, short_tag, 'Psi4', qmethod, qbasis))))
# Get omega conformer number of first, for reference info
# whole list can be used for matching purposes
indices_orig = pt.get_sd_list(imol, "original index")
# find conformers for which job did not finish (nan)
nanIndices = np.argwhere(np.isnan(iabs))
# convert energies from Hartrees to kcal/mol
iabs = 627.5095 * iabs
titleMols.append(imol.GetTitle())
confNans.append(nanIndices)
confNums.append(indices_orig)
enes.append(iabs)
return titleMols, confNums, enes, confNans
def avg_mol_time(titles, infile, method, basis, tag, mol_slice=[]):
"""
For an SDF file with all confs of all mols, get the average runtime
of all conformers for each molecule.
The input dictionary may or not be empty. If it is, append the avg/stdev
time of the calculation from this infile to existing molecule's value.
Parameters
----------
titles : dictionary
Keys are molecule names.
Values are [[qm1_avg, qm1_std], [qm2_avg, qm2_std], ... ]
where the index refers to a particular level of theory.
Dictionary may or may not be empty.
infile : string
name of the SDF file from which to extract time data from SD tag
method : string
QM method
basis : string
QM basis set
tag : string
datum of interest, e.g., "QM opt energy"
See keys in the define_tag function of proc_tags module.
mol_slice : list
list of indices from which to slice mols generator for read_mols
[start, stop, step]
Returns
-------
titles : dictionary
dictionary with extracted data from SDF file; keys are molnames,
values are lists of list of [avg_time, stdev_time] for many QM methods
"""
# Open molecule file.
if len(mol_slice) == 3:
mols = reader.read_mols(infile, mol_slice)
else:
mols = reader.read_mols(infile)
# Prepare text file to write extracted data.
timeF = open("timeAvgs.txt", 'a')
timeF.write("\nFile: {}\n".format(infile))
timeF.write(
"Average [{}/{}] [{}s] over all confs for each molecule\n".format(
method, basis, tag))
for mol_i in mols:
# get array of all conformer data of this mol
try:
time_array = np.fromiter(pt.get_sd_list(mol_i, tag, 'Psi4', method,
basis),
dtype=np.float64)
except ValueError:
time_array = np.asarray([np.nan]) * mol_i.NumConfs()
# exclude conformers for which job did not finish (nan)
nanIndices = np.argwhere(np.isnan(time_array))
for i in reversed(nanIndices): # loop in reverse to delete correctly
time_array = np.delete(time_array, i)
meantime = np.mean(time_array)
stdtime = np.std(time_array)
# write out data to file and store in dictionary
timeF.write(" %s\t%d confs\t\t%.3f +- %.3f\n" %
(mol_i.GetTitle(), time_array.size, meantime, stdtime))
name = mol_i.GetTitle()
if name not in titles:
titles[name] = []
titles[name].append([meantime, stdtime])
timeF.close()
return titles
def combine_files_plot(infile,
figname='combined.png',
molname=None,
verbose=False,
take_relative=False,
har_to_kcal=False):
"""
TODO
This only supports plotting of ONE specified molecule across different files.
Note on take_relative:
[1] Subtracting global minimum (single value) from all energies
doesn't work since everything is still on different scale.
subtract: (1) first conformer of each?, (2) global minimum?, (3) minimum of each?
Parameters
----------
infile : str
Filename with information on the files to read in, and
the SDF tags to be extracted from each. Columns are:
(1) QM method/basis, (2) sdf file, (3) tag key in sdf (like 'QM spe'),
(4) arbitrary label for plotting. Separate columns by comma.
molname
verbose
"""
wholedict = reader.read_text_input(infile)
numFiles = len(wholedict)
xarray = []
yarray = []
labels = []
titles = []
for i in wholedict:
print("Reading molecule(s) from file: ", wholedict[i]['fname'])
mols = reader.read_mols(wholedict[i]['fname'])
qmethod, qbasis = reader.separated_theory(wholedict[i]['theory'])
short_tag = wholedict[i]['tagkey']
for j, mol_j in enumerate(mols):
if molname is not None and mol_j.GetTitle() != molname:
continue
data_array = np.array(
list(
map(
float,
pt.get_sd_list(mol_j, short_tag, 'Psi4', qmethod,
qbasis))))
if take_relative:
data_array = data_array - data_array[0]
#data_array = data_array/data_array[0]
if har_to_kcal:
data_array = 627.5095 * data_array
titles.append(mol_j.GetTitle())
labels.append(wholedict[i]['label'])
yarray.append(data_array)
xarray.append(range(len(data_array)))
if verbose:
header = '{}\n'.format(molname)
for l in labels:
header += ("%s\n" % l)
xydata = np.vstack((xarray[0], yarray)).T
np.savetxt('combined.dat',
xydata,
delimiter='\t',
header=header,
fmt=' '.join(['%i'] + ['%10.4f'] * numFiles))
# letter labels for x-axis
num_confs = len(xarray[0])
letters = 'ABCDEFGHIJKLMNOPQRSTUVWXYZ'
rpt = int((num_confs / 26) + 1)
xlabs = [''.join(i)
for i in itertools.product(letters, repeat=rpt)][:num_confs]
fig = plt.figure()
ax = fig.add_subplot(111)
xlabel = 'conformer'
ylabel = "energy"
# vtl print max range of relative energies
conf_then_file = np.array(yarray).T
ranges = []
for c in conf_then_file:
c_spread = max(c) - min(c)
ranges.append(c_spread)
print(f'mol {molname} max range: {max(ranges)}')
ax.set_prop_cycle(
plt.cycler('color', plt.cm.rainbow(np.linspace(0, 1, len(yarray)))))
for i, (xs, ys) in enumerate(zip(xarray, yarray)):
plt.plot(xs, ys, '-o', lw=0.8, label=labels[i])
# publication view
# plt.ylabel(ylabel,fontsize=8)
# plt.xlabel(xlabel,fontsize=8)
# plt.legend(bbox_to_anchor=(0.08,1.05),loc=3,fontsize=8)
# fig.set_size_inches(3.37,1.7)
# standard view
plt.ylabel(ylabel, fontsize=14)
plt.xlabel(xlabel, fontsize=14)
plt.xticks(list(range(num_confs)), xlabs)
plt.legend(bbox_to_anchor=(1.05, 1), loc=2)
plt.title(molname)
plt.grid()
plt.savefig(figname, bbox_inches='tight', dpi=300)
plt.show()
def getRMSD(sdfRef, theory, rmsdict, package='Psi4'):
"""
Perform RMSD calculation from an SDF file for molecule and its conformers.
sdfRef: string, pathname of the SDF file with energies of opt 1 and opt 2
theory: string, level of theory in format of mp2/6-31G*
rmsdict: dictionary (can be empty) which will be populated in form of
rmsdict[theory][molName] = 0.000 if the RMSD of before/after energies are 0.000
package: string, name of software package used for QM calculation. only Psi4 currently supported
"""
method, basis = theory.split('/')[0].strip(), theory.split('/')[1].strip()
# create a molecule read in stream
print("Opening SDF file %s" % sdfRef)
molsRef = reader.read_mols(sdfRef)
# create file object for output RMSD calculation
RMSD = open("RMSD.txt", 'a')
RMSD.write("\nAnalyzing file: %s\n# Level of theory: %s\n" %
(sdfRef, theory))
# create file object for initial and final energies
energies = open("energies_breakdown.txt", 'a')
maximum = open("maxenergies.txt", "a")
# Grab energies, perform RMSD calculation, write data to txt files.
for rmol in molsRef:
molName = rmol.GetTitle()
tmol = np.asarray(pt.get_sd_list(rmol, 'QM opt energy', 'Psi4', method,
basis),
dtype=float)
imol = np.asarray(pt.get_sd_list(rmol, 'QM opt energy initial', 'Psi4',
method, basis),
dtype=float)
final = tmol.copy()
initial = imol.copy()
# subtract conformer[0] energies from all conformers
try:
tmol -= tmol[0]
except IndexError as e:
sys.exit("No energies found for {} {}/{}! Check that data is \
stored in tags. Exiting.".format(rmol.GetTitle(), method, basis))
imol -= imol[0]
#subtracts initial minus final and sqaures all values
fmol = np.subtract(tmol, imol)
fmol = fmol[~np.isnan(fmol)]
fmol = np.square(fmol)
#sums all energies of conformers for given rmol and then takes average with respect to n-1 number of conformers
tot = 0
for n in fmol:
tot += n
average = math.sqrt(tot / (fmol.size - 1))
#convert average from Hartree to Kcal/mol
average = average * 627.5095
# puts RMSD values into .txt file, and store in dict for plotting.
RMSD.write("#%s\t%.5f RMSD(Kcal/mol)\n" % (molName, average))
rmsdict[theory][molName] = average
# store energies of initial and final for molecules conformers in energies.txt
energies.write(
"\n#%s\n#%s\n#RMSD = %.5f(y)\t\t(x=Hartree, y=kcal/mol)\n#conf. init. Energy(x) \t final Energy(x) \t diff.(x)\tdiff. (y) \n"
% (theory, molName, average))
# get list of conformer indices to identify high RMSD ones
conflist = pt.get_sd_list(rmol, "original index", package, method,
basis)
conformer = []
for item in conflist:
conformer.append(item.split(',')[0]) # append orig conf
conformer = np.asarray(conformer, dtype=int)
difference = np.array([])
for i in range(len(tmol)):
energies.write(
"%r \t %5.9f \t %5.9f \t %5.9f\t%5.9f \n" %
(conformer[i], initial[i], final[i], final[i] - initial[i],
(final[i] - initial[i]) * 627.5095))
difference = np.append(difference,
[(final[i] - initial[i]) * 627.5095])
# find max 3 confs with highest RMSDs
try:
difference = np.absolute(difference)
confmax1 = (np.nanargmax(difference))
# set max conf to zero to find next highest
difference[confmax1] = 0
difference = np.absolute(difference)
confmax2 = (np.nanargmax(difference))
difference[confmax2] = 0
difference = np.absolute(difference)
confmax3 = (np.nanargmax(difference))
difference[confmax3] = 0
max1 = conformer[confmax1]
max2 = conformer[confmax2]
max3 = conformer[confmax3]
except ValueError as e:
#print("ValueError: {}".format(e))
# TODO don't plot this mol for all nan's
print("All RMSDs in list for file {} mol {} are nan!!!".format(
sdfRef, molName))
max1 = max2 = max3 = -1
energies.write(
"#*** Max energy differences are conformers (hi-->low): %r, %r, %r ***\n\n"
% (max1, max2, max3))
maximum.write("%s, %s : %r, %r, %r\n" %
(theory, molName, max1, max2, max3))
maximum.close()
RMSD.close()
energies.close()
return rmsdict
def get_psi_results(origsdf,
finsdf,
calctype='opt',
psiout="output.dat",
timeout="timer.dat"):
"""
Read in OEMols (and each of their conformers) in origsdf file,
get results from Psi4 calculations in the same directory as origsdf,
and write out results into finsdf file.
Directory layout is .../maindir/molName/confNumber/outputfiles .
Both origsdf and finsdf are located in maindir.
Parameters
----------
origsdf: string - original SDF file of input structures of QM calculation
finsdf: string - full name of final SDF file with optimized results.
calctype: string; one of 'opt','spe','hess' for geometry optimization,
single point energy calculation, or Hessian calculation
psiout: string - name of the Psi4 output files. Default is "output.dat"
timeout: string - name of the Psi4 timer files. Default is "timer.dat"
Returns
-------
method: string - QM method from Psi4 calculations
basisset: string - QM basis set from Psi4 calculations
None is returned if the function returns early (e.g., if output file
already exists)
"""
hdir, fname = os.path.split(origsdf)
wdir = os.getcwd()
# check that specified calctype is valid
if calctype not in {'opt', 'spe', 'hess'}:
raise ValueError("Specify a valid calculation type.")
# read in molecules
molecules = reader.read_mols(origsdf)
# open outstream file
writeout = os.path.join(wdir, finsdf)
write_ofs = oechem.oemolostream()
if os.path.exists(writeout):
raise FileExistsError(f"File already exists: {finsdf}\n")
if not write_ofs.open(writeout):
oechem.OEThrow.Fatal("Unable to open %s for writing" % writeout)
# Hessian dictionary, where hdict['molTitle']['confIndex'] has np array
if calctype == 'hess':
hdict = {}
# for each conformer, process output file and write new data to SDF file
for mol in molecules:
print("===== %s =====" % (mol.GetTitle()))
if calctype == 'hess':
hdict[mol.GetTitle()] = {}
for j, conf in enumerate(mol.GetConfs()):
props = initiate_dict()
# set file locations
timef = os.path.join(hdir,
"%s/%s/%s" % (mol.GetTitle(), j + 1, timeout))
outf = os.path.join(hdir,
"%s/%s/%s" % (mol.GetTitle(), j + 1, psiout))
# process output and get dictionary results
props = get_conf_data(props, calctype, timef, outf)
# if output was missing or are missing calculation details
# move on to next conformer
if props['missing'] or (calctype == 'opt' and not all(
key in props
for key in ['numSteps', 'finalEnergy', 'coords'])):
print(f"ERROR reading {outf}\nEither Psi4 job was incomplete "
"or wrong calctype specified\n")
method = None
basisset = None
continue
# add data to oemol
conf = set_conf_data(conf, props, calctype)
method = props['method']
basisset = props['basis']
# if hessian, append to dict bc does not go to SD tag
if calctype == 'hess':
hdict[mol.GetTitle()][j + 1] = props['hessian']
# check mol title
conf = check_title(conf, origsdf)
# write output file
oechem.OEWriteConstMolecule(write_ofs, conf)
# if hessian, write hdict out to separate file
if calctype == 'hess':
hfile = os.path.join(wdir,
os.path.splitext(finsdf)[0] + '.hess.pickle')
pickle.dump(hdict, open(hfile, 'wb'))
# close file streams
write_ofs.close()
return method, basisset