def test_single_deletion(self):
cobra_model = self.model
initialize_growth_medium(cobra_model, 'LB')
#Expected growth rates for the salmonella model with deletions in LB medium
the_loci = ['STM4081', 'STM0247', 'STM3867', 'STM2952']
the_genes = tpiA, metN, atpA, eno = map(cobra_model.genes.get_by_id, the_loci)
id_to_name = dict([(x.id, x.name) for x in the_genes])
growth_dict = {'fba':{tpiA.id:2.41, metN.id:2.44, atpA.id:1.87, eno.id:1.81},
'moma':{ tpiA.id:1.62, metN.id:2.4, atpA.id:1.40, eno.id:0.33}}
#MOMA requires cplex or gurobi
if get_solver_name(qp=True) is None:
growth_dict.pop('moma')
for method, the_growth_rates in growth_dict.items():
element_list = the_growth_rates.keys()
results = single_deletion(cobra_model, element_list=element_list,
method=method)
rates = results[0]
statuses = results[1]
for the_gene in element_list:
self.assertEqual(statuses[the_gene], 'optimal')
self.assertAlmostEqual(rates[the_gene], the_growth_rates[the_gene],
places=2)
def test_single_reaction_deletion(self):
cobra_model = self.model
results, status = single_deletion(
cobra_model, element_list=cobra_model.reactions[100:110:2],
element_type="reaction")
self.assertEqual(len(results), 5)
self.assertEqual(len(status), 5)
for status_value in status.values():
self.assertEqual(status_value, "optimal")
expected_results = {'3OAS140': 0, '3OAS160': 0.38001,
'3OAS180': 0.38001, '3OAS60': 0,
'3PEPTabcpp': 0.38001}
for reaction, value in results.items():
self.assertAlmostEqual(value, expected_results[reaction], 5)
def test_single_deletion(self):
cobra_model = self.model
initialize_growth_medium(cobra_model, 'LB')
#Expected growth rates for the salmonella model with deletions in LB medium
the_loci = ['STM4081', 'STM0247', 'STM3867', 'STM2952']
the_genes = tpiA, metN, atpA, eno = map(cobra_model.genes.get_by_id,
the_loci)
id_to_name = dict([(x.id, x.name) for x in the_genes])
growth_dict = {
'fba': {
tpiA.id: 2.41,
metN.id: 2.44,
atpA.id: 1.87,
eno.id: 1.81
},
'moma': {
tpiA.id: 1.62,
metN.id: 2.4,
atpA.id: 1.40,
eno.id: 0.33
}
}
#MOMA requires cplex or gurobi
try:
get_solver_name(qp=True)
except:
growth_dict.pop('moma')
for method, the_growth_rates in growth_dict.items():
element_list = the_growth_rates.keys()
results = single_deletion(cobra_model,
element_list=element_list,
method=method)
rates = results[0]
statuses = results[1]
for the_gene in element_list:
self.assertEqual(statuses[the_gene], 'optimal')
self.assertAlmostEqual(rates[the_gene],
the_growth_rates[the_gene],
places=2)
def deletion_analysis(
cobra_model,
the_medium=None,
deletion_type="single",
work_directory=None,
growth_cutoff=0.001,
the_problem="return",
n_processes=6,
element_type="gene",
solver="glpk",
error_reporting=None,
method="fba",
element_list=None,
):
"""Performs single and/or double deletion analysis on all the genes in the model. Provides
an interface to parallelize the deletion studies.
cobra_model: A Model object.
the_medium: Is None, a string, or a dictionary. If a string then the
initialize_growth_medium function expects that cobra_model has an
attribute dictionary called media_compositions, which is a dictionary of
dictionaries for various medium compositions. Where a medium
composition is a dictionary of exchange reaction ids for the medium
components and the exchange fluxes for each medium component; note that
these fluxes must be negative because they are being exchanged into the
system.
deletion_type: 'single', 'double', or 'double-only'
work_directory: None or String indicating where to save the output from the simulations.
growth_cutoff: Float. Indiates the minimum growth rate that is considered viable.
the_problem: Is None, 'return', or an LP model object for the solver.
element_type: 'gene' or 'reaction'
solver: 'glpk', 'gurobi', or 'cplex'
n_processes: number of parallel processes to break the double deletion simulations
into.
error_reporting: None or True
element_list: None or a list of genes to delete from the model.
Returns: Nothing. However, the script will add attributes single_deletion_* and
double_deletion_* to cobra_model containing the simulation results.
"""
if element_type == "reaction":
warn("deletion_analysis is not perfect for element_type = 'reaction'")
# When using ppmap, it's easier to feed in the parameters as a list,
# if the defaults need to be changed
if isinstance(cobra_model, list):
tmp_model = cobra_model
cobra_model = tmp_model[0]
if len(tmp_model) > 1:
the_medium = tmp_model[1]
if len(tmp_model) > 2:
deletion_type = tmp_model[2]
if len(tmp_model) > 3:
work_directory = tmp_model[3]
if len(tmp_model) > 4:
growth_cutoff = tmp_model[4]
if the_medium is not None:
initialize_growth_medium(cobra_model, the_medium)
the_problem = cobra_model.optimize(the_problem=the_problem, solver=solver)
# Store the basal model for the simulations
if element_list is None:
element_list = getattr(cobra_model, element_type + "s")
if deletion_type != "double_only":
cobra_model.single_deletion_growth_wt = cobra_model.solution.f
growth_rate_dict, growth_solution_status_dict, problem_dict = single_deletion(
deepcopy(cobra_model),
element_list=element_list,
the_problem=the_problem,
element_type=element_type,
solver=solver,
error_reporting=error_reporting,
method=method,
)
del problem_dict
cobra_model.single_deletion_growth_dict = growth_rate_dict
cobra_model.single_deletion_solution_status_dict = growth_solution_status_dict
setattr(cobra_model, "single_deletion_%ss" % element_type, deepcopy(growth_rate_dict.keys()))
cobra_model.single_deletion_lethal = [x for x in growth_rate_dict.keys() if growth_rate_dict[x] < growth_cutoff]
cobra_model.single_deletion_growth_medium = the_medium
cobra_model.single_deletion_nonlethal = list(
set(growth_rate_dict.keys()).difference(cobra_model.single_deletion_lethal)
)
if work_directory is not None:
if not path.lexists(work_directory):
mkdir(work_directory)
with open(work_directory + the_medium + "_single_" + cobra_model.description, "w") as out_file:
dump(cobra_model, out_file)
if deletion_type == "double" or deletion_type == "double_only":
# It appears that the glpk interface no longer works will with sending
# a glpk.LPX object through ppmap, so just set the basis to return
if the_problem:
the_problem = "return"
cobra_model.double_deletion_growth_medium = the_medium
cobra_model.double_deletion_growth_wt = cobra_model.solution.f
if not __parallel_mode_available:
if n_processes > 0:
print "Couldn't import ppmap from cobra.external is parallel python installed?"
return
else:
cobra_model = double_deletion_parallel(
deepcopy(cobra_model),
genes_of_interest=element_list,
the_problem=the_problem,
n_processes=n_processes,
element_type=element_type,
solver=solver,
error_reporting=error_reporting,
method=method,
)
# This indicates the genes that were run through double deletion but
# the x and y lists specify the order
setattr(cobra_model, "double_deletion_%ss" % element_type, deepcopy(cobra_model.genes))
if work_directory is not None:
with open(work_directory + the_medium + "_double_" + cobra_model.description, "w") as out_file:
dump(cobra_model, out_file)
return cobra_model
cobra_model = create_test_model(salmonella_pickle)
initialize_growth_medium(cobra_model, 'LB')
target_genes = ['STM4081', 'STM0247', 'STM3867', 'STM2952']
# Expected growth rates for the salmonella model after a deletions in LB medium
expected_growth_rates = {
"STM4081": 2.41,
"STM0247": 2.43,
"STM3867": 1.87,
"STM2952": 1.81}
start_time = time() # start timer
# Perform deletions for all genes in the list
rates, statuses = single_deletion(cobra_model, target_genes)
total_time = time() - start_time # stop timer
# print out results
passed_string = 'PASSED: %s simulation (%1.3f) ~= expectation (%1.2f)'
failed_string = 'FAILED: %s simulation (%1.3f) != expectation (%1.2f)'
for gene_locus, rate in rates.items():
# get gene name from gene locus (i.e. STM4081 -> tpiA)
name = cobra_model.genes.get_by_id(gene_locus).name
# test if the simulation failed
if statuses[gene_locus] != "optimal":
print("deletion failed for %s (%s)" % (name, gene_locus))
if abs(rate - expected_growth_rates[gene_locus]) > 0.01:
print(failed_string % (name, rate, expected_growth_rates[gene_locus]))
else:
def double_gene_deletion_moma(cobra_model, gene_list_1=None, gene_list_2=None,
method='moma', single_deletion_growth_dict=None,
solver='glpk', growth_tolerance=1e-8,
error_reporting=None):
"""This will disable reactions for all gene pairs from gene_list_1 and
gene_list_2 and then run simulations to optimize for the objective
function. The contribution of each reaction to the objective function
is indicated in cobra_model.reactions[:].objective_coefficient vector.
NOTE: We've assumed that there is no such thing as a synthetic rescue with
this modeling framework.
cobra_model: a cobra.Model object
gene_list_1: Is None or a list of genes. If None then both gene_list_1
and gene_list_2 are assumed to correspond to cobra_model.genes.
gene_list_2: Is None or a list of genes. If None then gene_list_2 is
assumed to correspond to gene_list_1.
method: 'fba' or 'moma' to run flux balance analysis or minimization
of metabolic adjustments.
single_deletion_growth_dict: A dictionary that provides the growth
rate information for single gene knock outs. This can speed up
simulations because nonviable single deletion strains imply that all
double deletion strains will also be nonviable.
solver: 'glpk', 'gurobi', or 'cplex'.
error_reporting: None or True
growth_tolerance: float. The effective lower bound on the growth rate
for a single deletion that is still considered capable of growth.
Returns a dictionary of the gene ids in the x dimension (x) and the y
dimension (y), and the growth simulation data (data).
"""
#BUG: Since this might be called from ppmap, the modules need to
#be imported. Modify ppmap to take depfuncs
from numpy import zeros
nan = float('nan')
from cobra.flux_analysis.single_deletion import single_deletion
from cobra.manipulation import delete_model_genes, undelete_model_genes
##TODO: Use keywords instead
if isinstance(cobra_model, dict):
tmp_dict = cobra_model
cobra_model = tmp_dict['cobra_model']
if 'gene_list_1' in tmp_dict:
gene_list_1 = tmp_dict['gene_list_1']
if 'gene_list_2' in tmp_dict:
gene_list_2 = tmp_dict['gene_list_2']
if 'method' in tmp_dict:
method = tmp_dict['method']
if 'single_deletion_growth_dict' in tmp_dict:
single_deletion_growth_dict = tmp_dict['single_deletion_growth_dict']
if 'solver' in tmp_dict:
solver = tmp_dict['solver']
if 'error_reporting' in tmp_dict:
error_reporting = tmp_dict['error_reporting']
else:
cobra_model = cobra_model
#this is a slow way to revert models.
wt_model = cobra_model #NOTE: It may no longer be necessary to use a wt_model
#due to undelete_model_genes
if gene_list_1 is None:
gene_list_1 = cobra_model.genes
elif not hasattr(gene_list_1[0], 'id'):
gene_list_1 = map(cobra_model.genes.get_by_id, gene_list_1)
#Get default values to use if the deletions do not alter any reactions
cobra_model.optimize(solver=solver)
basal_f = cobra_model.solution.f
if method.lower() == 'moma':
wt_model = cobra_model.copy()
combined_model = None
single_gene_set = set(gene_list_1)
if gene_list_2 is not None:
if not hasattr(gene_list_2[0], 'id'):
gene_list_2 = map(cobra_model.genes.get_by_id, gene_list_2)
single_gene_set.update(gene_list_2)
#Run the single deletion analysis to account for double deletions that
#target the same gene and lethal deletions. We assume that there
#aren't synthetic rescues.
single_deletion_growth_dict = single_deletion(cobra_model,
list(single_gene_set),
method=method,
solver=solver)[0]
if gene_list_2 is None or gene_list_1 == gene_list_2:
number_of_genes = len(gene_list_1)
gene_list_2 = gene_list_1
deletion_array = zeros([number_of_genes, number_of_genes])
##TODO: Speed up this triangular process
#For the case where the contents of the lists are the same cut the work in half.
#There might be a faster way to do this by using a triangular array function
#in numpy
#Populate the diagonal from the single deletion lists
for i, the_gene in enumerate(gene_list_1):
deletion_array[i, i] = single_deletion_growth_dict[the_gene.id]
for i, gene_1 in enumerate(gene_list_1[:-1]):
#TODO: Since there cannot be synthetic rescues we can assume
#that the whole row for a lethal deletion
#will be equal to that deletion.
if single_deletion_growth_dict[gene_1.id] < growth_tolerance:
tmp_solution = single_deletion_growth_dict[gene_1.id]
for j in range(i+1, number_of_genes):
deletion_array[j, i] = deletion_array[i, j] = tmp_solution
else:
for j, gene_2 in enumerate(gene_list_1[i+1:], i+1):
if single_deletion_growth_dict[gene_2.id] < growth_tolerance:
tmp_solution = single_deletion_growth_dict[gene_2.id]
else:
delete_model_genes(cobra_model, [gene_1, gene_2])
if cobra_model._trimmed:
if method.lower() == 'fba':
#Assumes that the majority of perturbations don't change
#reactions which is probably false
cobra_model.optimize(solver=solver, error_reporting=error_reporting)
the_status = cobra_model.solution.status
tmp_solution = cobra_model.solution.f
elif method.lower() == 'moma':
try:
moma_solution = moma(wt_model, cobra_model,
combined_model=combined_model,
solver=solver)
tmp_solution = float(moma_solution.pop('objective_value'))
the_status = moma_solution.pop('status')
combined_model = moma_solution.pop('combined_model')
del moma_solution
except:
tmp_solution = nan
the_status = 'failed'
if the_status not in ['opt', 'optimal'] and \
error_reporting:
print('%s / %s: %s status: %s'%(gene_1, gene_2, solver,
the_status))
#Reset the model to orginial form.
undelete_model_genes(cobra_model)
else:
tmp_solution = basal_f
deletion_array[j, i] = deletion_array[i, j] = tmp_solution
else:
deletion_array = zeros([len(gene_list_1), len(gene_list_2)])
#Now deal with the case where the gene lists are different
for i, gene_1 in enumerate(gene_list_1):
if single_deletion_growth_dict[gene_1.id] <= 0:
for j in range(len(gene_list_2)):
deletion_array[i, j] = 0.
else:
for j, gene_2 in enumerate(gene_list_2):
#Assume no such thing as a synthetic rescue
if single_deletion_growth_dict[gene_2.id] <= growth_tolerance:
tmp_solution = single_deletion_growth_dict[gene_2.id]
else:
delete_model_genes(cobra_model, [gene_1, gene_2])
if cobra_model._trimmed:
if method.lower() == 'fba':
cobra_model.optimize(solver=solver)
tmp_solution = cobra_model.solution.f
the_status = cobra_model.solution.status
elif method.lower() == 'moma':
try:
moma_solution = moma(wt_model, cobra_model,
combined_model=combined_model,
solver=solver)
tmp_solution = float(moma_solution.pop('objective_value'))
the_status = moma_solution.pop('status')
combined_model = moma_solution.pop('combined_model')
del moma_solution
except:
tmp_solution = nan
the_status = 'failed'
if the_status not in ['opt', 'optimal'] and \
error_reporting:
print('%s / %s: %s status: %s'%(repr(gene_1), repr(gene_2), solver,
cobra_model.solution.status))
#Reset the model to wt form
undelete_model_genes(cobra_model)
else:
tmp_solution = basal_f
deletion_array[i, j] = tmp_solution
if hasattr(gene_list_1, 'id'):
gene_list_1 = [x.id for x in gene_list_1]
if hasattr(gene_list_2, 'id'):
gene_list_2 = [x.id for x in gene_list_2]
return({'x': gene_list_1, 'y': gene_list_2, 'data': deletion_array})
def double_gene_deletion(
cobra_model,
gene_list_1=None,
gene_list_2=None,
method="fba",
single_deletion_growth_dict=None,
the_problem="return",
solver="glpk",
error_reporting=None,
):
"""This will disable reactions for all gene pairs from gene_list_1 and
gene_list_2 and then run simulations to optimize for the objective
function. The contribution of each reaction to the objective function
is indicated in cobra_model.reactions[:].objective_coefficient vector.
cobra_model: a cobra.Model object
gene_list_1: Is None or a list of genes. If None then both gene_list_1
and gene_list_2 are assumed to correspond to cobra_model.genes.
gene_list_2: Is None or a list of genes. If None then gene_list_2 is
assumed to correspond to gene_list_1.
method: 'fba' or 'moma' to run flux balance analysis or minimization
of metabolic adjustments.
single_deletion_growth_dict: A dictionary that provides the growth
rate information for single gene knock outs. This can speed up
simulations because nonviable single deletion strains imply that all
double deletion strains will also be nonviable.
the_problem: Is None, 'return', or an LP model object for the solver.
solver: 'glpk', 'gurobi', or 'cplex'.
error_reporting: None or True
Returns a dictionary of the genes in the x dimension (x), the y
dimension (y), and the growth simulation data (data).
"""
# BUG: Since this might be called from ppmap, the modules need to
# be imported. Modify ppmap to take depfuncs
from numpy import zeros, nan
from cobra.flux_analysis.single_deletion import single_deletion
from cobra.manipulation import initialize_growth_medium
from cobra.manipulation import delete_model_genes, undelete_model_genes
##TODO: Use keywords instead
if isinstance(cobra_model, dict):
tmp_dict = cobra_model
cobra_model = tmp_dict["cobra_model"]
if "gene_list_1" in tmp_dict:
gene_list_1 = tmp_dict["gene_list_1"]
if "gene_list_2" in tmp_dict:
gene_list_2 = tmp_dict["gene_list_2"]
if "method" in tmp_dict:
method = tmp_dict["method"]
if "the_problem" in tmp_dict:
the_problem = tmp_dict["the_problem"]
if "single_deletion_growth_dict" in tmp_dict:
single_deletion_growth_dict = tmp_dict["single_deletion_growth_dict"]
if "solver" in tmp_dict:
solver = tmp_dict["solver"]
if "error_reporting" in tmp_dict:
error_reporting = tmp_dict["error_reporting"]
else:
cobra_model = cobra_model
# this is a slow way to revert models.
wt_model = cobra_model # NOTE: It may no longer be necessary to use a wt_model
# due to undelete_model_genes
if gene_list_1 is None:
gene_list_1 = cobra_model.genes
elif not hasattr(gene_list_1[0], "id"):
gene_list_1 = map(cobra_model.genes.get_by_id, gene_list_1)
# Get default values to use if the deletions do not alter any reactions
the_problem = cobra_model.optimize(the_problem=the_problem, solver=solver)
basal_f = cobra_model.solution.f
if method.lower() == "moma":
wt_model = cobra_model.copy()
the_problem = "return"
combined_model = None
single_gene_set = set(gene_list_1)
if gene_list_2:
if not hasattr(gene_list_2[0], "id"):
gene_list_2 = map(cobra_model.genes.get_by_id, gene_list_2)
single_gene_set.update(gene_list_2)
# Run the single deletion analysis to account for double deletions that
# target the same gene and lethal deletions. We assume that there
# aren't synthetic rescues.
single_deletion_growth_dict = single_deletion(
cobra_model,
list(single_gene_set),
method=method,
the_problem=the_problem,
solver=solver,
error_reporting=error_reporting,
)[0]
if gene_list_2 is None or gene_list_1 == gene_list_2:
deletion_array = zeros([len(gene_list_1), len(gene_list_1)])
##TODO: Speed up this triangular process
# For the case where the contents of the lists are the same cut the work in half.
# There might be a faster way to do this by using a triangular array function
# in numpy
# Populate the diagonal from the single deletion lists
for i, the_gene in enumerate(gene_list_1):
deletion_array[i, i] = single_deletion_growth_dict[the_gene.id]
for i in range(len(gene_list_1) - 1):
gene_1 = gene_list_1[i]
# TODO: Since there cannot be synthetic rescues we can assume
# that the whole row for a lethal deletion
# will be equal to that deletion.
if single_deletion_growth_dict[gene_1.id] <= 0:
for j in range(i + 1, len(gene_list_1)):
deletion_array[j, i] = deletion_array[i, j] = single_deletion_growth_dict[gene_2.id]
else:
for j in range(i + 1, len(gene_list_1)):
if single_deletion_growth_dict[gene_1.id] <= 0:
tmp_solution = single_deletion_growth_dict[gene_1.id]
else:
gene_2 = gene_list_1[j]
delete_model_genes(cobra_model, [gene_1, gene_2])
if cobra_model._trimmed:
if method.lower() == "fba":
# Assumes that the majority of perturbations don't change
# reactions which is probably false
cobra_model.optimize(
the_problem=the_problem, solver=solver, error_reporting=error_reporting
)
the_status = cobra_model.solution.status
tmp_solution = cobra_model.solution.f
elif method.lower() == "moma":
try:
moma_solution = moma(
wt_model,
cobra_model,
combined_model=combined_model,
solver=solver,
the_problem=the_problem,
)
tmp_solution = float(moma_solution.pop("objective_value"))
the_problem = moma_solution.pop("the_problem")
the_status = moma_solution.pop("status")
combined_model = moma_solution.pop("combined_model")
del moma_solution
except:
tmp_solution = nan
the_status = "failed"
if the_status not in ["opt", "optimal"] and error_reporting:
print "%s / %s: %s status: %s" % (gene_1, gene_2, solver, the_status)
# Reset the model to orginial form.
undelete_model_genes(cobra_model)
else:
tmp_solution = basal_f
deletion_array[j, i] = deletion_array[i, j] = tmp_solution
else:
deletion_array = zeros([len(gene_list_1), len(gene_list_2)])
# Now deal with the case where the gene lists are different
for i, gene_1 in enumerate(gene_list_1):
if single_deletion_growth_dict[gene_1.id] <= 0:
for j in range(len(gene_list_2)):
deletion_array[i, j] = 0.0
else:
for j, gene_2 in enumerate(gene_list_2):
# Assume no such thing as a synthetic rescue
if single_deletion_growth_dict[gene_2.id] <= 0:
tmp_solution = single_deletion_growth_dict[gene_2.id]
else:
delete_model_genes(cobra_model, [gene_1, gene_2])
if cobra_model._trimmed:
if method.lower() == "fba":
cobra_model.optimize(
the_problem=the_problem, solver=solver, error_reporting=error_reporting
)
tmp_solution = cobra_model.solution.f
the_status = cobra_model.solution.status
elif method.lower() == "moma":
try:
moma_solution = moma(
wt_model,
cobra_model,
combined_model=combined_model,
solver=solver,
the_problem=the_problem,
)
tmp_solution = float(moma_solution.pop("objective_value"))
the_problem = moma_solution.pop("the_problem")
the_status = moma_solution.pop("status")
combined_model = moma_solution.pop("combined_model")
del moma_solution
except:
tmp_solution = nan
the_status = "failed"
if the_status not in ["opt", "optimal"] and error_reporting:
print "%s / %s: %s status: %s" % (
repr(gene_1),
repr(gene_2),
solver,
cobra_model.solution.status,
)
# Reset the model to wt form
undelete_model_genes(cobra_model)
else:
tmp_solution = basal_f
deletion_array[i, j] = tmp_solution
return {"x": gene_list_1, "y": gene_list_2, "data": deletion_array}
def double_gene_deletion_moma(cobra_model,
gene_list_1=None,
gene_list_2=None,
method='moma',
single_deletion_growth_dict=None,
solver='glpk',
growth_tolerance=1e-8,
error_reporting=None):
"""This will disable reactions for all gene pairs from gene_list_1 and
gene_list_2 and then run simulations to optimize for the objective
function. The contribution of each reaction to the objective function
is indicated in cobra_model.reactions[:].objective_coefficient vector.
NOTE: We've assumed that there is no such thing as a synthetic rescue with
this modeling framework.
cobra_model: a cobra.Model object
gene_list_1: Is None or a list of genes. If None then both gene_list_1
and gene_list_2 are assumed to correspond to cobra_model.genes.
gene_list_2: Is None or a list of genes. If None then gene_list_2 is
assumed to correspond to gene_list_1.
method: 'fba' or 'moma' to run flux balance analysis or minimization
of metabolic adjustments.
single_deletion_growth_dict: A dictionary that provides the growth
rate information for single gene knock outs. This can speed up
simulations because nonviable single deletion strains imply that all
double deletion strains will also be nonviable.
solver: 'glpk', 'gurobi', or 'cplex'.
error_reporting: None or True
growth_tolerance: float. The effective lower bound on the growth rate
for a single deletion that is still considered capable of growth.
Returns a dictionary of the gene ids in the x dimension (x) and the y
dimension (y), and the growth simulation data (data).
"""
#BUG: Since this might be called from ppmap, the modules need to
#be imported. Modify ppmap to take depfuncs
from numpy import zeros
nan = float('nan')
from cobra.flux_analysis.single_deletion import single_deletion
from cobra.manipulation import delete_model_genes, undelete_model_genes
##TODO: Use keywords instead
if isinstance(cobra_model, dict):
tmp_dict = cobra_model
cobra_model = tmp_dict['cobra_model']
if 'gene_list_1' in tmp_dict:
gene_list_1 = tmp_dict['gene_list_1']
if 'gene_list_2' in tmp_dict:
gene_list_2 = tmp_dict['gene_list_2']
if 'method' in tmp_dict:
method = tmp_dict['method']
if 'single_deletion_growth_dict' in tmp_dict:
single_deletion_growth_dict = tmp_dict[
'single_deletion_growth_dict']
if 'solver' in tmp_dict:
solver = tmp_dict['solver']
if 'error_reporting' in tmp_dict:
error_reporting = tmp_dict['error_reporting']
else:
cobra_model = cobra_model
#this is a slow way to revert models.
wt_model = cobra_model #NOTE: It may no longer be necessary to use a wt_model
#due to undelete_model_genes
if gene_list_1 is None:
gene_list_1 = cobra_model.genes
elif not hasattr(gene_list_1[0], 'id'):
gene_list_1 = map(cobra_model.genes.get_by_id, gene_list_1)
#Get default values to use if the deletions do not alter any reactions
cobra_model.optimize(solver=solver)
basal_f = cobra_model.solution.f
if method.lower() == 'moma':
wt_model = cobra_model.copy()
combined_model = None
single_gene_set = set(gene_list_1)
if gene_list_2 is not None:
if not hasattr(gene_list_2[0], 'id'):
gene_list_2 = map(cobra_model.genes.get_by_id, gene_list_2)
single_gene_set.update(gene_list_2)
#Run the single deletion analysis to account for double deletions that
#target the same gene and lethal deletions. We assume that there
#aren't synthetic rescues.
single_deletion_growth_dict = single_deletion(cobra_model,
list(single_gene_set),
method=method,
solver=solver)[0]
if gene_list_2 is None or gene_list_1 == gene_list_2:
number_of_genes = len(gene_list_1)
gene_list_2 = gene_list_1
deletion_array = zeros([number_of_genes, number_of_genes])
##TODO: Speed up this triangular process
#For the case where the contents of the lists are the same cut the work in half.
#There might be a faster way to do this by using a triangular array function
#in numpy
#Populate the diagonal from the single deletion lists
for i, the_gene in enumerate(gene_list_1):
deletion_array[i, i] = single_deletion_growth_dict[the_gene.id]
for i, gene_1 in enumerate(gene_list_1[:-1]):
#TODO: Since there cannot be synthetic rescues we can assume
#that the whole row for a lethal deletion
#will be equal to that deletion.
if single_deletion_growth_dict[gene_1.id] < growth_tolerance:
tmp_solution = single_deletion_growth_dict[gene_1.id]
for j in range(i + 1, number_of_genes):
deletion_array[j, i] = deletion_array[i, j] = tmp_solution
else:
for j, gene_2 in enumerate(gene_list_1[i + 1:], i + 1):
if single_deletion_growth_dict[
gene_2.id] < growth_tolerance:
tmp_solution = single_deletion_growth_dict[gene_2.id]
else:
delete_model_genes(cobra_model, [gene_1, gene_2])
if cobra_model._trimmed:
if method.lower() == 'fba':
#Assumes that the majority of perturbations don't change
#reactions which is probably false
cobra_model.optimize(
solver=solver,
error_reporting=error_reporting)
the_status = cobra_model.solution.status
tmp_solution = cobra_model.solution.f
elif method.lower() == 'moma':
try:
moma_solution = moma(
wt_model,
cobra_model,
combined_model=combined_model,
solver=solver)
tmp_solution = float(
moma_solution.pop('objective_value'))
the_status = moma_solution.pop('status')
combined_model = moma_solution.pop(
'combined_model')
del moma_solution
except:
tmp_solution = nan
the_status = 'failed'
if the_status not in ['opt', 'optimal'] and \
error_reporting:
print('%s / %s: %s status: %s' %
(gene_1, gene_2, solver, the_status))
#Reset the model to orginial form.
undelete_model_genes(cobra_model)
else:
tmp_solution = basal_f
deletion_array[j, i] = deletion_array[i, j] = tmp_solution
else:
deletion_array = zeros([len(gene_list_1), len(gene_list_2)])
#Now deal with the case where the gene lists are different
for i, gene_1 in enumerate(gene_list_1):
if single_deletion_growth_dict[gene_1.id] <= 0:
for j in range(len(gene_list_2)):
deletion_array[i, j] = 0.
else:
for j, gene_2 in enumerate(gene_list_2):
#Assume no such thing as a synthetic rescue
if single_deletion_growth_dict[
gene_2.id] <= growth_tolerance:
tmp_solution = single_deletion_growth_dict[gene_2.id]
else:
delete_model_genes(cobra_model, [gene_1, gene_2])
if cobra_model._trimmed:
if method.lower() == 'fba':
cobra_model.optimize(solver=solver)
tmp_solution = cobra_model.solution.f
the_status = cobra_model.solution.status
elif method.lower() == 'moma':
try:
moma_solution = moma(
wt_model,
cobra_model,
combined_model=combined_model,
solver=solver)
tmp_solution = float(
moma_solution.pop('objective_value'))
the_status = moma_solution.pop('status')
combined_model = moma_solution.pop(
'combined_model')
del moma_solution
except:
tmp_solution = nan
the_status = 'failed'
if the_status not in ['opt', 'optimal'] and \
error_reporting:
print('%s / %s: %s status: %s' %
(repr(gene_1), repr(gene_2), solver,
cobra_model.solution.status))
#Reset the model to wt form
undelete_model_genes(cobra_model)
else:
tmp_solution = basal_f
deletion_array[i, j] = tmp_solution
if hasattr(gene_list_1, 'id'):
gene_list_1 = [x.id for x in gene_list_1]
if hasattr(gene_list_2, 'id'):
gene_list_2 = [x.id for x in gene_list_2]
return ({'x': gene_list_1, 'y': gene_list_2, 'data': deletion_array})
cobra_model = create_test_model(salmonella_pickle)
initialize_growth_medium(cobra_model, 'LB')
target_genes = ['STM4081', 'STM0247', 'STM3867', 'STM2952']
# Expected growth rates for the salmonella model after a deletions in LB medium
expected_growth_rates = {
"STM4081": 2.41,
"STM0247": 2.43,
"STM3867": 1.87,
"STM2952": 1.81
}
start_time = time() # start timer
# Perform deletions for all genes in the list
rates, statuses = single_deletion(cobra_model, target_genes)
total_time = time() - start_time # stop timer
# print out results
passed_string = 'PASSED: %s simulation (%1.3f) ~= expectation (%1.2f)'
failed_string = 'FAILED: %s simulation (%1.3f) != expectation (%1.2f)'
for gene_locus, rate in rates.items():
# get gene name from gene locus (i.e. STM4081 -> tpiA)
name = cobra_model.genes.get_by_id(gene_locus).name
# test if the simulation failed
if statuses[gene_locus] != "optimal":
print("deletion failed for %s (%s)" % (name, gene_locus))
if abs(rate - expected_growth_rates[gene_locus]) > 0.01:
print(failed_string % (name, rate, expected_growth_rates[gene_locus]))
else:
