def on_run(self):
# Build Workflow
wf = Workflow(project_dir='./loop_workflow')
wf.task_runner = self
# Task 1: Task that will provide an array of values that will be
# iterated over. Setting output_format=mol2 directly as input
# to this task is to demonstrate that parameters passed to the
# LoopTask will be forwarded to the mapped workflows created by
# the LoopTask. Equally so for steps=100 illustrating that input
# of steps in the workflow can obtained from tasks outside of
# the loop
t1 = wf.add_task('Array provider')
t1.set_input(output_format='mol2', steps=100)
# Task 2: Add loop task. The 'mapper_arg' defines the parameter name in
# the input that holds an iterable of input values to iterate
# over. The 'loop_end_task' is required and defines the task
# that 'closes' the loop and collects all results.
t2 = wf.add_task('Loop',
task_type='LoopTask',
mapper_arg='smiles',
loop_end_task='Collector')
wf.connect_task(t1.nid, t2.nid)
# Task 3: Convert SMILES to mol2
# Convert ligand to mol2 format irrespective of input format.
t3 = wf.add_task('Ligand conversion',
task_type='WampTask',
uri='mdgroup.mdstudio_structures.endpoint.convert')
wf.connect_task(t2.nid, t3.nid, smiles='mol')
# Task 4: Convert mol2 to 3D mol2 irrespective if input is 1D/2D or 3D
# mol2 If 'output_format' is not specified it is deduced from
# the input wich is mol2 in this case. There are circumstances
# where conversion to 3D fails, retry upto 3 times.
t4 = wf.add_task('Make_3D',
task_type='WampTask',
uri='mdgroup.mdstudio_structures.endpoint.make3d',
retry_count=3)
wf.connect_task(t3.nid, t4.nid, 'mol')
wf.connect_task(t1.nid, t4.nid, 'steps')
# Task 5: Empty task that server as a collector for all results
# obtained during iteration.
t5 = wf.add_task('Collector')
wf.connect_task(t4.nid, t5.nid, 'mol')
# Set the array of input SMILES string to task 1
wf.input(t1.nid,
smiles=[
'O1[C@@H](CCC1=O)CCC',
'C[C@]12CC[C@H]3[C@@H](CC=C4CCCC[C@]34CO)[C@@H]1CCC2=O',
'CC12CCC3C(CC=C4C=CCCC34C)C1CCC2=O'
])
wf.run()
while wf.is_running:
yield sleep(1)
def on_run(self):
"""
When the microservice has successfully registered with the broker the
on_run method is the first method to be called.
We are using this method now to run our example workflow.
"""
# Workflow constants, these will be saved as part of the workflow
# specification
ligand_format = 'smi'
pH = 7.4
protein_file = os.path.abspath('protein.mol2')
protein_binding_center = [4.9264, 19.0796, 21.9892]
# Build Workflow
wf = Workflow(description='MDStudio WAMP workflow')
# Task 1: convert the SMILES string to mol2 format (2D).
# Add a task using the 'add_task' method always defining:
# an administrative title of the task and the task type here a WampTask
# because we are calling an microservice endpoint defined by uri.
# 'store_output' is True by default and stores the task input/output to disk.
t1 = wf.add_task('Format_conversion',
task_type='WampTask',
uri='mdgroup.mdstudio_structures.endpoint.convert')
# Use 'set_input' do define the input to a task. As we are now building
# a workflow specification these will be task constants but the same
# method will be used later on to define specific input when using the
# workflow specification for a ligand.
t1.set_input(output_format='mol2')
# Task 2: Covert mol2 to 3D mol2 irrespective if input is 1D/2D or 3D mol2
# This particular 3D conversion routine is known to fail sometimes but by
# setting retry_count to 3 the workflow manager will retry 3 times before
# failing.
t2 = wf.add_task('Make_3D',
task_type='WampTask',
uri='mdgroup.mdstudio_structures.endpoint.make3d',
retry_count=3)
t2.set_input(output_format='mol2')
# Use 'connect_task' to connect t1 to t2 using their unique identifiers
# (nid). In addition we can specify the parameters for task 1 we wish to
# use as input to task 2 as additional argument or keyword arguments to
# the functions. A keyword argument defines a parameter name mapping
# between the two tasks.
wf.connect_task(t1.nid, t2.nid, 'mol')
# Task 3: Adjust ligand protonation state to a given pH if applicable
t3 = wf.add_task('Add hydrogens',
task_type='WampTask',
uri='mdgroup.mdstudio_structures.endpoint.addh')
t3.set_input(output_format='mol2', correctForPH=True, pH=pH)
wf.connect_task(t2.nid, t3.nid, 'mol')
# Task 4: Get the formal charge for the protonated mol2 to use as input
# for ACPYPE or ATB
# Here store_output equals False which will keep all output in memory and
# finally as part of the stored workflow file (*.jgf)
t4 = wf.add_task('Get charge',
task_type='WampTask',
uri='mdgroup.mdstudio_structures.endpoint.info',
store_output=False)
t4.set_input(input_format='mol2')
wf.connect_task(t3.nid, t4.nid, 'mol')
# Task 5: Create rotations of the molecule for better sampling
t5 = wf.add_task('Create 3D rotations',
task_type='WampTask',
uri='mdgroup.mdstudio_structures.endpoint.rotate')
t5.set_input(rotations=[[1, 0, 0, 90], [1, 0, 0, -90], [0, 1, 0, 90],
[0, 1, 0, -90], [0, 0, 1, 90], [0, 0, 1, -90]])
wf.connect_task(t3.nid, t5.nid, 'mol')
# Task 6: Run PLANTS on ligand and protein
# The 'workdir' argument points to a tmp directory that is shared between
# the microservice docker image and the host system to store results.
t6 = wf.add_task('Plants docking',
task_type='WampTask',
uri='mdgroup.mdstudio_smartcyp.endpoint.docking')
t6.set_input(cluster_structures=100,
bindingsite_center=protein_binding_center,
bindingsite_radius=12,
protein_file=protein_file,
threshold=3.0,
base_work_dir='/tmp/mdstudio/mdstudio_smartcyp')
# Here we pass only the 'mol' parameter from task 5 to task 6 where it is
# named 'ligand_file'
wf.connect_task(t5.nid, t6.nid, 'mol', mol='ligand_file')
# Task 7: Extract cluster medians from output using a custom function.
# A task of type 'PythonTask' allows to add custom python functions
# or classes to the workflow. They are defined using the 'custom_func'
# parameter according to the Python import syntax. The package or file
# containing the function should be available as part of the PYTHONPATH.
t7 = wf.add_task('Get cluster medians',
task_type='PythonTask',
custom_func='workflow_helpers.get_docking_medians')
wf.connect_task(t6.nid, t7.nid, 'result')
# Task 8: retrieve median structures
t8 = wf.add_task(
'Retrieve median structures',
task_type='WampTask',
uri='mdgroup.mdstudio_smartcyp.endpoint.docking_structures')
t8.set_input(create_ensemble=False)
wf.connect_task(t7.nid, t8.nid, medians='paths')
# Save the workflow specification
wf.save('workflow_spec.jgf')
# Lets run the workflow specification for a number of ligand SMILES
# The current microservice instance (self) is passed as task_runner to the workflow
# it will be used to make calls to other microservice endpoints when task_type equals WampTask.
wf.task_runner = self
currdir = os.getcwd()
for i, ligand in enumerate([
'O1[C@@H](CCC1=O)CCC',
'C[C@]12CC[C@H]3[C@@H](CC=C4CCCC[C@]34CO)[C@@H]1CCC2=O',
'CC12CCC3C(CC=C4C=CCCC34C)C1CCC2=O'
],
start=1):
wf.load('workflow_spec.jgf')
wf.input(t1.nid,
mol={
'content': ligand,
'path': None,
'extension': ligand_format
})
wf.run(project_dir='./ligand-{0}'.format(i))
while wf.is_running:
yield sleep(1)
os.chdir(currdir)
def on_run(self):
# Ligand to make prediction for
ligand = 'O1[C@@H](CCC1=O)CCC'
ligand_format = 'smi'
liemodel = os.path.join(os.getcwd(), '1A2_model')
# CYP1A2 pre-calibrated model
modelpicklefile = os.path.join(liemodel, 'params.pkl')
modelfile = pickle.load(open(modelpicklefile))
unbound_trajectory = os.path.join(os.getcwd(),
"unbound_trajectory.ene")
bound_trajectory = [os.path.join(os.getcwd(), "bound_trajectory.ene")]
decompose_files = [
os.path.join(os.getcwd(), "decompose_dataframe.ene")
]
# Build Workflow
wf = Workflow(project_dir='./lie_prediction')
wf.task_runner = self
# STAGE 5. PYLIE FILTERING, AD ANALYSIS AND BINDING-AFFINITY PREDICTION
# Collect Gromacs bound and unbound MD energy trajectories in a dataframe
t18 = wf.add_task(
'Create mdframe',
task_type='WampTask',
uri='mdgroup.lie_pylie.endpoint.collect_energy_trajectories')
t18.set_input(unbound_trajectory=unbound_trajectory,
bound_trajectory=bound_trajectory,
lie_vdw_header="Ligand-Ligenv-vdw",
lie_ele_header="Ligand-Ligenv-ele")
# Determine stable regions in MDFrame and filter
t19 = wf.add_task(
'Detect stable regions',
task_type='WampTask',
uri='mdgroup.lie_pylie.endpoint.filter_stable_trajectory')
t19.set_input(do_plot=True,
minlength=45,
workdir='/tmp/mdstudio/lie_pylie')
wf.connect_task(t18.nid, t19.nid, 'mdframe')
# Extract average LIE energy values from the trajectory
t20 = wf.add_task(
'LIE averages',
task_type='WampTask',
uri='mdgroup.lie_pylie.endpoint.calculate_lie_average')
wf.connect_task(t19.nid, t20.nid, filtered_mdframe='mdframe')
# Calculate dG using pre-calibrated model parameters
t21 = wf.add_task('Calc dG',
task_type='WampTask',
uri='mdgroup.lie_pylie.endpoint.liedeltag')
t21.set_input(alpha_beta_gamma=modelfile['LIE']['params'])
wf.connect_task(t20.nid, t21.nid, 'averaged', averaged='dataframe')
# Applicability domain: 1. Tanimoto similarity with training set
t22 = wf.add_task(
'AD1 tanimoto simmilarity',
task_type='WampTask',
uri='mdgroup.mdstudio_structures.endpoint.chemical_similarity')
t22.set_input(test_set=[ligand],
mol_format=ligand_format,
reference_set=modelfile['AD']['Tanimoto']['smi'],
ci_cutoff=modelfile['AD']['Tanimoto']['Furthest'])
wf.connect_task(t18.nid, t22.nid)
# Applicability domain: 2. residue decomposition
t23 = wf.add_task('AD2 residue decomposition',
task_type='WampTask',
uri='mdgroup.lie_pylie.endpoint.adan_residue_decomp',
inline_files=False)
t23.set_input(model_pkl=modelpicklefile,
decompose_files=decompose_files)
wf.connect_task(t18.nid, t23.nid)
# Applicability domain: 3. deltaG energy range
t24 = wf.add_task('AD3 dene yrange',
task_type='WampTask',
uri='mdgroup.lie_pylie.endpoint.adan_dene_yrange')
t24.set_input(ymin=modelfile['AD']['Yrange']['min'],
ymax=modelfile['AD']['Yrange']['max'])
wf.connect_task(t21.nid,
t24.nid,
'liedeltag_file',
liedeltag_file='dataframe')
# Applicability domain: 4. deltaG energy distribution
t25 = wf.add_task('AD4 dene distribution',
task_type='WampTask',
uri='mdgroup.lie_pylie.endpoint.adan_dene')
t25.set_input(model_pkl=modelpicklefile,
center=list(modelfile['AD']['Dene']['Xmean']),
ci_cutoff=modelfile['AD']['Dene']['Maxdist'])
wf.connect_task(t21.nid,
t25.nid,
'liedeltag_file',
liedeltag_file='dataframe')
wf.run()
while wf.is_running:
yield sleep(1)
def on_run(self):
# Ligand to make prediction for
ligand = 'O1[C@@H](CCC1=O)CCC'
ligand_format = 'smi'
liemodel = os.path.join(os.getcwd(), '1A2_model')
# CYP1A2 Model data
with open(os.path.join(liemodel, 'model.dat'), 'r') as mdf:
model = json.load(mdf)
# CYP1A2 pre-calibrated model
modelpicklefile = os.path.join(liemodel, 'params.pkl')
modelfile = pickle.load(open(modelpicklefile))
# Build Workflow
wf = Workflow(project_dir='./allies_run')
wf.task_runner = self
# STAGE 1: LIGAND PRE-PROCESSING
# Convert ligand to mol2 irrespective of input format.
t1 = wf.add_task('Format_conversion',
task_type='WampTask',
uri='mdgroup.mdstudio_structures.endpoint.convert')
t1.set_input(mol={
'content': ligand,
'path': None,
'extension': ligand_format
})
# Covert mol2 to 3D mol2 irrespective if input is 1D/2D or 3D mol2
t2 = wf.add_task('Make_3D',
task_type='WampTask',
uri='mdgroup.mdstudio_structures.endpoint.make3d')
t2.set_input(output_format='mol2')
wf.connect_task(t1.nid, t2.nid, 'mol')
# Adjust ligand protonation state to a given pH if applicable
t3 = wf.add_task('Add hydrogens',
task_type='WampTask',
uri='mdgroup.mdstudio_structures.endpoint.addh')
t3.set_input(output_format='mol2',
correctForPH=model['pHCorr'],
pH=model['pH'])
wf.connect_task(t2.nid, t3.nid, 'mol')
# Get the formal charge for the protonated mol2 to use as input for ACPYPE or ATB
t4 = wf.add_task('Get charge',
task_type='WampTask',
uri='mdgroup.mdstudio_structures.endpoint.info')
t4.set_input(input_format='mol2')
wf.connect_task(t3.nid, t4.nid, 'mol')
# # STAGE 2. CREATE TOPOLOGY FOR LIGAND
# Run acpype on ligands
t5 = wf.add_task('ACPYPE',
task_type='WampTask',
uri='mdgroup.mdstudio_amber.endpoint.acpype',
retry_count=3)
wf.connect_task(t3.nid, t5.nid, mol='structure')
wf.connect_task(t4.nid, t5.nid, charge='net_charge')
# STAGE 3. PLANTS DOCKING
# Create rotations of the molecule for better sampling
t6 = wf.add_task('Create 3D rotations',
task_type='WampTask',
uri='mdgroup.mdstudio_structures.endpoint.rotate')
t6.set_input(rotations=[[1, 0, 0, 90], [1, 0, 0, -90], [0, 1, 0, 90],
[0, 1, 0, -90], [0, 0, 1, 90], [0, 0, 1, -90]])
wf.connect_task(t3.nid, t6.nid, 'mol')
# Run PLANTS on ligand and protein
t7 = wf.add_task('Plants docking',
task_type='WampTask',
uri='mdgroup.mdstudio_smartcyp.endpoint.docking')
t7.set_input(cluster_structures=100,
bindingsite_center=model['proteinParams'][0]['pocket'],
bindingsite_radius=model['proteinParams'][0]['radius'],
protein_file=os.path.join(
liemodel, model['proteinParams'][0]['proteinDock']),
threshold=3.0,
base_work_dir='/tmp/mdstudio/mdstudio_smartcyp')
wf.connect_task(t6.nid, t7.nid, mol='ligand_file')
# Get cluster median structures from docking
t8 = wf.add_task(
'Get cluster medians',
custom_func='allies_workflow_helpers.get_docking_medians')
wf.connect_task(t7.nid, t8.nid, 'output')
# STAGE 4. GROMACS MD
# Ligand in solution
t14 = wf.add_task(
'MD ligand in water',
task_type='WampTask',
uri='mdgroup.mdstudio_gromacs.endpoint.gromacs_ligand',
store_output=True)
t14.set_input(
sim_time=0.001, #sim_time=model['timeSim'],
include=[
os.path.join(liemodel, model['proteinTopPos']),
os.path.join(liemodel, 'attype.itp')
],
residues=model['resSite'],
protein_file=None,
protein_top=os.path.join(liemodel, model['proteinTop']),
cerise_file=os.path.join(os.getcwd(), 'cerise_config_gt.json'))
wf.connect_task(t5.nid,
t14.nid,
new_pdb='ligand_file',
gmx_itp='topology_file')
# convert PLANTS mol2 to pdb
t15 = wf.add_task('Ligand mol2 to PDB',
task_type='WampTask',
uri='mdgroup.mdstudio_structures.endpoint.convert')
t15.set_input(output_format='pdb')
wf.connect_task(t8.nid, t15.nid, medians='mol')
# Run MD for protein + ligand
t16 = wf.add_task(
'MD protein-ligand',
task_type='WampTask',
uri='mdgroup.mdstudio_gromacs.endpoint.gromacs_protein')
t16.set_input(sim_time=0.001,
include=[
os.path.join(liemodel, model['proteinTopPos']),
os.path.join(liemodel, 'attype.itp')
],
residues=model['resSite'],
charge=model['charge'],
cerise_file=os.path.join(os.getcwd(),
'cerise_config_gt.json'),
protein_file=os.path.join(
liemodel, model['proteinParams'][0]['proteinCoor']),
protein_top=os.path.join(liemodel, model['proteinTop']))
wf.connect_task(t15.nid, t16.nid, mol='ligand_file')
wf.connect_task(t5.nid, t16.nid, gmx_itp='topology_file')
# Collect results
t17 = wf.add_task(
'Collect MD results',
custom_func='allies_workflow_helpers.collect_md_enefiles')
t17.set_input(model_dir=liemodel)
wf.connect_task(t14.nid, t17.nid, output='unbound')
wf.connect_task(t16.nid, t17.nid, output='bound')
# STAGE 5. PYLIE FILTERING, AD ANALYSIS AND BINDING-AFFINITY PREDICTION
# Collect Gromacs bound and unbound MD energy trajectories in a dataframe
t18 = wf.add_task(
'Create mdframe',
task_type='WampTask',
uri='mdgroup.lie_pylie.endpoint.collect_energy_trajectories')
t18.set_input(lie_vdw_header="Ligand-Ligenv-vdw",
lie_ele_header="Ligand-Ligenv-ele")
wf.connect_task(t17.nid, t18.nid, 'bound_trajectory',
'unbound_trajectory')
# Determine stable regions in MDFrame and filter
t19 = wf.add_task(
'Detect stable regions',
task_type='WampTask',
uri='mdgroup.lie_pylie.endpoint.filter_stable_trajectory')
t19.set_input(do_plot=True, FilterSplines={'minlength': 45})
wf.connect_task(t18.nid, t19.nid, 'mdframe')
# Extract average LIE energy values from the trajectory
t20 = wf.add_task(
'LIE averages',
task_type='WampTask',
uri='mdgroup.lie_pylie.endpoint.calculate_lie_average')
wf.connect_task(t19.nid, t20.nid, filtered_mdframe='mdframe')
# Calculate dG using pre-calibrated model parameters
t21 = wf.add_task('Calc dG',
task_type='WampTask',
uri='mdgroup.lie_pylie.endpoint.liedeltag')
t21.set_input(alpha=modelfile['LIE']['params'][0],
beta=modelfile['LIE']['params'][1],
gamma=modelfile['LIE']['params'][2])
wf.connect_task(t20.nid, t21.nid, averaged='dataframe')
# Applicability domain: 1. Tanimoto similarity with training set
t22 = wf.add_task(
'AD1 tanimoto simmilarity',
task_type='WampTask',
uri='mdgroup.lie_structures.endpoint.chemical_similarity')
t22.set_input(test_set=[ligand],
mol_format=ligand_format,
reference_set=modelfile['AD']['Tanimoto']['smi'],
ci_cutoff=modelfile['AD']['Tanimoto']['Furthest'])
wf.connect_task(t18.nid, t22.nid)
# Applicability domain: 2. residue decomposition
t23 = wf.add_task('AD2 residue decomposition',
task_type='WampTask',
uri='mdgroup.lie_pylie.endpoint.adan_residue_decomp')
t23.set_input(model_pkl=modelpicklefile)
wf.connect_task(t17.nid, t23.nid, 'decomp_files')
# Applicability domain: 3. deltaG energy range
t24 = wf.add_task('AD3 dene yrange',
task_type='WampTask',
uri='mdgroup.lie_pylie.endpoint.adan_dene_yrange')
t24.set_input(ymin=modelfile['AD']['Yrange']['min'],
ymax=modelfile['AD']['Yrange']['max'])
wf.connect_task(t21.nid, t24.nid, liedeltag_file='dataframe')
# Applicability domain: 4. deltaG energy distribution
t25 = wf.add_task('AD4 dene distribution',
task_type='WampTask',
uri='mdgroup.lie_pylie.endpoint.adan_dene')
t25.set_input(model_pkl=modelpicklefile,
center=list(modelfile['AD']['Dene']['Xmean']),
ci_cutoff=modelfile['AD']['Dene']['Maxdist'])
wf.connect_task(t21.nid, t25.nid, liedeltag_file='dataframe')
# Save the workflow specification
wf.save('workflow_spec.jgf')
wf.run()
while wf.is_running:
yield sleep(1)