def convert(subdir):
"""Convert united atom files to all atom format."""
# Get substition numbers from table
hydrogen_per_atom = read_atom_numbers()
with cd(subdir):
pdb_path = find(path='..',
folder_keyword='initial',
file_keyword='.pdb')[0]
top_path = find(path='..',
folder_keyword='initial',
file_keyword='.top')[0]
for dft_type in ('esp', 'rho'):
dft_path = find(path='..',
folder_keyword='dft_calculations',
file_keyword='{}.cube'.format(dft_type))[0]
out_file = dft_type + '_ua' + '.cube'
kwargs = {
'infile_pdb': pdb_path,
'infile_top': top_path,
'infile_cube': dft_path,
'outfile_cube': out_file,
'implicitHbondingPartners': hydrogen_per_atom
}
# Call Johannes' conversion script
smamp.aa2ua_cube.aa2ua_cube(**kwargs)
def main():
""" Run the script."""
print('This is {}.'.format(__file__))
# Create a folder for the averaged cost function
chargepath = './horton_charges/sweep_rhoref'
if os.path.isdir(chargepath):
pass
# shutil.rmtree(chargepath)
else:
os.makedirs(chargepath)
cost_paths = find(path='.',
folder_keyword='4_horton_cost_function/lnrho_sweep',
file_keyword='cost',
nr_occ=None)
lnrho_range = []
sigma_range = []
for charge_file in cost_paths:
# print(charge_file)
# Parse parameters from filename
lnrho, sigma = charge_file[-15:-3].split('_')[-2:]
lnrho_range += [lnrho]
sigma_range += [sigma]
lnrho_range = set(lnrho_range)
sigma_range = set(sigma_range)
for lnrho in lnrho_range:
print('lnrho = {}'.format(lnrho))
for sigma in sigma_range:
filename = 'cost_{}_{}.h5'.format(lnrho, sigma)
cost_function_paths = find(path='.',
folder_keyword='lnrho_sweep',
file_keyword=filename,
nr_occ=None)
# print(cost_function_paths)
if len(cost_function_paths) == 0:
print('WARNING: No cost functions found! Continue ..')
continue
# Extract cost function As and Bs
A_matrices, B_vectors = collect_matrices(cost_function_paths)
# Average over all matrices & vectors
average_A, average_B = average(A_matrices, B_vectors)
# keep one HDF5 file as a template for writing into later
out_name = 'costfunction_average_{}_{}.h5'.format(lnrho, sigma)
template_path = os.path.join(chargepath, out_name)
shutil.copyfile(cost_function_paths[0], template_path)
# Export matrices to hdf5
export(average_A, average_B, template_path=template_path)
print('Done.')
def main():
""" Execute everything."""
print('This is {}.'.format(__file__))
rho_paths = find(path='./', folder_keyword='dft', file_keyword='rho')
for path in rho_paths:
# Get the exact density file name
folder_path, file_name = os.path.split(path)
# Go to the bader directory to place the bader analysis output there
topdir = os.path.split(folder_path)[0]
bader_dir = os.path.join(topdir, '5_bader_charges')
with cd(bader_dir):
print('Moving to {}'.format(bader_dir))
# Find structure and topology files of the same snapshot
snapshot_path = find(path='..', folder_keyword='initial',
file_keyword='.pdb')[0]
top_path = find(path='..', folder_keyword='initial',
file_keyword='.top')[0]
if os.path.exists('bader_charges.csv'):
print('File exists, skipping.')
continue
# Write output to logfile
with open('bader.log', 'w') as logfile:
# Assemble the shell command bader
command = 'bader -p atom_index '
command += os.path.join('../2_dft_calculations/',
file_name)
kwargs = {"shell": True, "stdout": logfile,
"stderr": subprocess.STDOUT}
# Execute the shell command
print('Running bader ...')
p = subprocess.Popen(command, **kwargs)
# Wait for the shell command to finish
p.communicate()
# Extract charges from the bader anaylsis output to .csv
print('Bader done. Extracting bader charges ...')
smamp.extract_bader_charges.extract(snapshot_path,
top_path)
print('Extraction done.')
print('Done.')
def collect_avg():
collect_df = pd.DataFrame()
cost_paths = find(path='.',
folder_keyword='horton_charges/sweep_rhoref',
file_keyword='charges',
nr_occ=None)
for charge_file in cost_paths:
# Parse parameters from filename
lnrho, sigma = charge_file[-15:-4].split('_')[-2:]
sarah_wish_sigma = ['0.2', '0.4', '0.6', '0.8', '1.0', '1.2', '1.4']
if sigma in sarah_wish_sigma:
# Read file
df = pd.read_csv(charge_file)
# Paste the lnrho parameter into the dataframe
df['lnrhoref'] = float(lnrho)
df['sigma'] = float(sigma)
# Also note the snapshot identifier
df['diff'] = (df.q - df.q_unconstrained)**2
collect_df = collect_df.append(df)
collect_df = pd.melt(collect_df,
id_vars=['atom', 'residue', 'lnrhoref', 'sigma'],
value_vars=['diff'])
return collect_df
def collect_snapshots(plot_range=[-9, -5]):
collect_df = pd.DataFrame()
cost_paths = find(path='.',
folder_keyword='4_horton_cost_function/lnrho_sweep',
file_keyword='charge',
nr_occ=None)
for charge_file in cost_paths:
# Parse parameters from filename
lnrho, sigma = charge_file[-15:-4].split('_')[-2:]
sarah_wish_sigma = ['0.2', '0.4', '0.6', '0.8', '1.0', '1.2', '1.4']
if sigma in sarah_wish_sigma:
# Read file
df = pd.read_csv(charge_file)
# Paste the lnrho parameter into the dataframe
df['lnrhoref'] = float(lnrho)
df['sigma'] = float(sigma)
# Also note the snapshot identifier
timestamp = re.findall(r'\d+', charge_file)[0]
df['snapshot'] = timestamp
# df['diff'] = (df.q - df.q_unconstrained).abs()
df['diff'] = (df.q - df.q_unconstrained)**2
collect_df = collect_df.append(df)
collect_df = pd.melt(
collect_df,
id_vars=['atom', 'residue', 'snapshot', 'lnrhoref', 'sigma'],
value_vars=['diff'])
return collect_df
def collect_snapshots():
cost_paths = find(path='.', folder_keyword='4_horton_cost_function/lnrho_sweep',
file_keyword='charges',
nr_occ=None)
collect_df = pd.DataFrame()
for charge_file in cost_paths:
# Parse parameters from filename
lnrho, sigma = charge_file[-15:-4].split('_')[-2:]
#if sigma == '0.8':
if sigma in ['0.2', '0.4', '0.6', '0.8', '1.0', '1.2', '1.4']:
df = pd.read_csv(charge_file)
# Paste the lnrho parameter into the dataframe
df['lnrho'] = lnrho
df['sigma'] = sigma
# Also note the snapshot identifier
timestamp = re.findall(r'\d+', charge_file)[0]
df['snapshot'] = timestamp
collect_df = collect_df.append(df)
collect_df = pd.melt(collect_df,
id_vars=['atom', 'residue', 'lnrho', 'sigma', 'snapshot'],
value_vars=['q'])
return collect_df
def collect_bfgs_energies():
"""Find BFGS energies and put them in one dataframe."""
# Initialize collection data frame
coll_df = None
# Crawl the directory structure
bfgs_list = find(path='./', folder_keyword='dft', file_keyword='BFGS', nr_occ=None)
for energy_file in bfgs_list:
print('Moving to {}'.format(energy_file))
topdir = os.path.split(os.path.split(energy_file)[0])[0]
time = topdir.replace('./', '').replace('_ps_snapshot', '')
time = int(time)
with open(energy_file) as data:
string = data.read()
string = re.sub(r"\[.+\]", "", string)
string = re.sub(r"\*", "", string)
string = re.sub('BFGSLineSearch: ', '', string)
sio = StringIO(string)
df = pd.read_csv(sio, sep='\s+', engine='python', skiprows=[1])
df['Timestamp'] = time
if coll_df is None:
coll_df = df
else:
coll_df = coll_df.append(df)
return coll_df
def collect_averages():
collect_df = pd.DataFrame()
cost_paths = find(path='.', folder_keyword='horton_charges/sweep_rhoref',
file_keyword='charge',
nr_occ=None)
for charge_file in cost_paths:
# Parse parameters from filename
lnrho, sigma = charge_file[-15:-4].split('_')[-2:]
if sigma == '0.8':
charge_file = find(path='.', folder_keyword='horton_charges/sweep_rhoref',
file_keyword='charges_{}_{}.csv'.format(lnrho, sigma),
nr_occ=1)[0]
df = pd.read_csv(charge_file)
df['lnrho'] = float(lnrho)
collect_df = collect_df.append(df)
collect_df = pd.melt(collect_df, id_vars=['atom', 'residue', 'lnrho'], value_vars=['q'])
return collect_df
def read_input_files():
"""Search for and read input files (with implicit H-atoms)."""
ase_struct, pmd_top = None, None
pdb_file = find(path='..',
folder_keyword='initial_structure',
file_keyword='.pdb')[0]
top_file = find(path='..',
folder_keyword='initial_structure',
file_keyword='.top')[0]
ase_struct = ase.io.read(pdb_file)
pmd_struct = pmd.load_file(pdb_file)
pmd_top = gromacs.GromacsTopologyFile(top_file, parametrize=False)
# Make sure we actually found everything we need
if ase_struct is None:
raise RuntimeError(
'structure file (.pdb) not found in {}'.format(input_dir))
if pmd_top is None:
raise RuntimeError(
'topology file (.top) not found in {}'.format(input_dir))
return ase_struct, pmd_struct, pmd_top
"""Visualize the convergence of potential differences.
Copyright 2019 Simulation Lab
University of Freiburg
Author: Lukas Elflein <*****@*****.**>
"""
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from smamp.tools import find
print('Load data (wide format) ...')
data_paths = find(path='.',
folder_keyword='/data',
file_keyword='csv',
nr_occ=None)
collect_df = pd.DataFrame()
for data_file in data_paths:
df = pd.read_csv(data_file)
collect_df = collect_df.append(df)
print('Manipulating data ...')
df = collect_df.reset_index(drop=True)
df.columns = df.columns.map(str.strip)
df['lnrho'] = pd.to_numeric(df['lnrho'])
df['sigma'] = df['sigma'].apply(lambda x: str(x))
df = df.loc[df.lnrho < -3]
df_0 = df.loc[df.charge == 0]
df_1 = df.loc[df.charge == 1]
def main():
print('This is {}.'.format(__file__))
charges = [0, 1, 2]
random.shuffle(charges)
for charge in charges:
pmd2ase = None
root = '../{}_charge/1_charge_cycle'.format(charge)
pdb_path = find(path=root,
folder_keyword='0_initial_structure',
file_keyword='.pdb')[0]
top_path = find(path=root,
folder_keyword='0_initial_structure',
file_keyword='.top')[0]
hyd_path = find(path='../{}_charge'.format(charge),
folder_keyword='fitting',
file_keyword='hydrogen_per_atom.csv')[0]
# Find all files with point charges
charge_paths = find(path=root,
folder_keyword='4_horton_cost_function/lnrho',
file_keyword='charges',
nr_occ=None)
lnrho_range = []
sigma_range = []
time_range = []
print('{} paths found.'.format(len(charge_paths)))
random.shuffle(charge_paths)
i = 0
for charge_path in charge_paths:
i += 1
if i % 10 == 0:
print('{} of {}'.format(i, len(charge_paths)))
work_dir = os.path.split(
os.path.split(os.path.split(charge_path)[0])[0])[0]
cube_path = find(path=work_dir,
folder_keyword='2_dft_calculations',
file_keyword='esp.cube')[0]
# Parse parameters from filename
lnrho, sigma = charge_path[-15:-4].split('_')[-2:]
time = charge_path.split('/')[3].split('_')[0]
if not sigma in [
'0.2', '0.4', '0.6', '0.8', '1.0', '1.2', '1.4', '1.6'
]:
print('skipping sigma {}'.format(sigma))
continue
out_path = 'data/pot_err_{}_{}_{}_{}.csv'.format(
charge, time, lnrho, sigma)
if os.path.exists(out_path):
print('exists, skipping.')
continue
# Calculate the error between DFT-ESP and point-charge ESP
rrmsd, pmd2ase = calc_error(cube_path,
pdb_path,
top_path,
hyd_path,
charge_path,
pmd2ase=pmd2ase,
n_samples=40000)
# Write to file
with open(out_path, 'w') as outfile:
outfile.write('charge, time, lnrho, sigma, rrmsd\n')
outfile.write('{}, {}, {}, {}, {}\n'.format(
charge, time, lnrho, sigma, rrmsd))
print('Done.')
def main():
""" Execute everything."""
print('This is {}.'.format(__file__))
# We need any one top and pdb file for the ordering of atom-names;
# The exact snapshot we use does not matter as the ordering is invariant
pdb_file = find(path='.',
folder_keyword='0_initial_structure',
file_keyword='.pdb')[0]
top_file = find(path='.',
folder_keyword='0_initial_structure',
file_keyword='.top')[0]
hyd_file = find(path='..',
folder_keyword='fitting',
file_keyword='hydrogen_per_atom.csv')[0]
cost_paths = find(path='.',
folder_keyword='4_horton_cost_function/lnrho_sweep',
file_keyword='cost',
nr_occ=None)
lnrho_range = []
sigma_range = []
for charge_file in cost_paths:
# Parse parameters from filename
lnrho, sigma = charge_file[-15:-3].split('_')[-2:]
lnrho_range += [lnrho]
sigma_range += [sigma]
lnrho_range = set(lnrho_range)
sigma_range = set(sigma_range)
for lnrho in lnrho_range:
print('lnrho = {} ...'.format(lnrho))
for sigma in sigma_range:
# Find the paths for the unaveraged snapshot cost functions
cost_paths = find(
path='.',
folder_keyword='4_horton_cost_function/lnrho_sweep',
file_keyword='cost_{}_{}.h5'.format(lnrho, sigma),
nr_occ=None)
# Find the path for the average cost function
cost_avg = find(
path='.',
folder_keyword='horton_charges/sweep_rhoref',
file_keyword='costfunction_average_{}_{}.h5'.format(
lnrho, sigma),
nr_occ=None)
cost_paths += cost_avg
for cost_file in cost_paths:
folder = os.path.split(cost_file)[0]
output_file = os.path.join(
folder, 'charges_{}_{}.csv'.format(lnrho, sigma))
if os.path.exists(output_file):
# print('{} exists. Skipping ahead.'.format(output_file))
continue
print('Optimizing charges for {}.'.format(cost_file[:18]))
calc_charges(pdb_file,
top_file,
hyd_file,
cost_file,
output_file=output_file)
print('Done.')
def calc_charges(pdb_infile,
top_infile,
hydrogen_file,
horton_cost_function,
output_file,
charge_group_file=None,
charge_group_charges_file=None,
symmetry_file=None):
'''Wraps fitESPconstrained.py'''
# Look up the relationship between ASE indices, atom names
pmd_struct, pmd_top, ase2pmd = create_structure(pdb_infile, top_infile,
hydrogen_file)
print('Atomic structure built.')
# Import A and B matrices from HORTON
A, B = read_horton_cost_function(horton_cost_function)
if charge_group_file is None:
try:
charge_group_file = find('..', 'fitting_constraint_files',
'atoms_in_charge_group.csv')[0]
except:
charge_group_file = None
if charge_group_charges_file is None:
try:
charge_group_charges_file = find(
'..', 'fitting_constraint_files',
'charge_group_total_charge.csv')[0]
except:
charge_group_charges_file = None
if symmetry_file is None:
try:
symmetry_file = find('..', 'fitting_constraint_files',
'atoms_of_same_charge.csv')[0]
except:
symmetry_file = None
# Calculate constraints
logic_constraints, charge_constraints = get_constraints(
args=None,
ase2pmd=ase2pmd,
charge_group_file=charge_group_file,
charge_group_charges_file=charge_group_charges_file,
symmetry_file=symmetry_file,
debug=False)
print('Constraints calculated: {} non-redunant.'.format(
logic_constraints.shape[0]))
# Run the constrained minimization
q, f = constrained_minimize(A, B, logic_constraints, charge_constraints)
# print('Constrained minimization done.')
# print('Extremal charges: {:1.5f}, {:1.5f}'.format(q.min(), q.max()))
# print('Extremal Lagrange forces: {:1.5f}, {:1.5f}'.format(f.min(), f.max()))
q_unconstrained = unconstrained_minimize(A, B)
# Save charges
charge_df = write_charges(q,
q_unconstrained,
ase2pmd,
out_name=output_file,
plot=False)
# Save Lagrange forces
# write_forces(f, logic_constraints, ase2pmd)
print('Charges written to {}.'.format(output_file))
def main():
""" Execute everything."""
print('This is {}.'.format(__file__))
# Save the working dir
topdir = os.getcwd()
print('Current working dir: {}'.format(topdir))
# Read the file containing the info on the total charge of the system
qtot = read_total_charge(
path='../fitting_constraint_files/total_charge.csv')
# Crawl the directory structure
for subdir, dirs, files in sorted(os.walk('.')):
# Exclude template folders from search
if 'template' in subdir or 'exclude' in subdir or 'sweep' in subdir:
continue
# Select the folders with cost function
if 'horton_cost_function' in subdir or 'horton_charges' in subdir:
print('Moving to {}'.format(subdir))
with cd(subdir):
# Search for the .pdb and .top file in the folders above
input_path = '..'
pdb_path = find(input_path,
folder_keyword='initial',
file_keyword='.pdb',
nr_occ=1,
exclude_kw=['template', 'exclude', 'sweep'])[0]
top_path = find(input_path,
folder_keyword='initial',
file_keyword='.top',
nr_occ=1,
exclude_kw=['template', 'exclude', 'sweep'])[0]
output_file = 'fitted_point_charges.csv'
cost_file = 'cost.h5'
constraint_path = '../../../fitting_constraint_files/'
if 'horton_charges' in subdir:
constraint_path = '../../fitting_constraint_files/'
cost_file = 'costfunction_average.h5'
hydrogen_file = os.path.join(constraint_path,
'hydrogen_per_atom.csv')
charge_group_file = os.path.join(constraint_path,
'atoms_in_charge_group.csv')
charge_group_charges_file = os.path.join(
constraint_path, 'charge_group_total_charge.csv')
symmetry_file = os.path.join(constraint_path,
'atoms_of_same_charge.csv')
if os.path.isfile(output_file):
print('File exists, skipping: {}'.format(
os.path.join(subdir, output_file)))
continue
calc_charges(
pdb_infile=pdb_path,
top_infile=top_path,
horton_cost_function=cost_file,
output_file=output_file,
hydrogen_file=hydrogen_file,
charge_group_file=charge_group_file,
charge_group_charges_file=charge_group_charges_file,
symmetry_file=symmetry_file)