def fva(self) -> pd.DataFrame:
model = self.read_model()
result = fba(model)
fluxes = result.fluxes
try:
fva_result = flux_variability_analysis(
model, reactions=model.reactions, fraction_of_optimum=1.0
) # type: FluxVariabilityResult
df = fva_result.data_frame
df_out = pd.DataFrame(
{
"model": self.model_path.name,
"objective": self.objective_id,
"reaction": df.index,
"flux": fluxes,
"status": CuratorConstants.STATUS_OPTIMAL,
"minimum": df.lower_bound,
"maximum": df.upper_bound,
}
)
except Exception as e:
logger.error(f"{e}")
df_out = pd.DataFrame(
{
"model": self.model_path.name,
"objective": self.objective_id,
"reaction": [r.id for r in model.reactions],
"flux": fluxes,
"status": CuratorConstants.STATUS_INFEASIBLE,
"minimum": CuratorConstants.VALUE_INFEASIBLE,
"maximum": CuratorConstants.VALUE_INFEASIBLE,
}
)
return df_out
def reaction_deletion(self) -> pd.DataFrame:
model = self.read_model()
reaction_status = []
reaction_values = []
for reaction in model.reactions:
reaction_bounds = (reaction.lower_bound, reaction.upper_bound)
reaction.bounds = (0, 0)
# run fba
try:
result = fba(model)
value = result.objective_value
status = CuratorConstants.STATUS_OPTIMAL
except Exception as e:
logger.error(f"{e}")
value = CuratorConstants.VALUE_INFEASIBLE
status = CuratorConstants.STATUS_INFEASIBLE
reaction_status.append(status)
reaction_values.append(value)
# restore bounds
reaction.bounds = reaction_bounds[:]
return pd.DataFrame(
{
"model": self.model_path.name,
"objective": self.objective_id,
"reaction": [r.id for r in model.reactions],
"status": reaction_status,
"value": reaction_values,
}
)
def optimize_models_cameo(model_paths: List[Path]) -> pd.DataFrame:
"""FBA optimization for all given models."""
results = []
n_models = len(model_paths)
for k, path in enumerate(model_paths):
# load model
start_time = time.time()
model = read_sbml_model(str(path))
load_time = time.time() - start_time # [s]
# run optimization
start_time = time.time()
result = fba(model)
simulate_time = time.time() - start_time # [s]
objective_value = result.objective_value
filename = path.name
model = filename.split(".")[0]
res = (model, objective_value, load_time, simulate_time)
results.append(res)
print("[{}/{}]".format(k, n_models), res)
return pd.DataFrame(data=results,
columns=("model", "objective_value", "load_time",
"simulate_time"))
def write_results():
"""optimize results and write out to a file"""
from cameo import load_model, fba
outF = open("KOFBAFluxes-2.txt","w")
model = load_model('KOmodel.json')
print(len(model.genes))
print(len(model.metabolites))
print(len(model.reactions))
# new_solution = model.optimize()
fba_result = fba(model)
print(fba_result)
# print(fba_result.data_frame)
print(len(fba_result.data_frame))
# print(model.reactions)
print(dir(fba_result))
print(fba_result.data_frame)
fbaD = fba_result.data_frame.T.to_dict()
fbaCorrD = {}
for keyS, valueF in fbaD.items():
fbaCorrD[keyS] = valueF["flux"]
for keyS, valueF in fbaCorrD.items():
# print(keyS, valueF)
outF.write(str(keyS) + ", " + str(valueF) + "\n")
def test_apply_designs(self, model, diff_fva):
result = diff_fva.run()
works = []
for strain_design in result:
with model:
strain_design.apply(model)
try:
solution = fba(model, objective="Biomass_Ecoli_core_N_lp_w_fsh_GAM_rp__Nmet2")
works.append(solution["EX_succ_lp_e_rp_"] > 1e-6 and solution.objective_value > 1e-6)
except Infeasible:
works.append(False)
assert any(works)
def evaluate_design(model, strain_design, pathway, aerobic, bpcy, pyield):
with TimeMachine() as tm:
if not aerobic and 'EX_o2_e' in model.reactions:
model.reactions.EX_o2_e.change_bounds(lb=0, time_machine=tm)
pathway.apply(model, time_machine=tm)
strain_design.apply(model, time_machine=tm)
try:
solution = fba(model, objective=model.biomass)
_bpcy = bpcy(model, solution, strain_design.targets)
_pyield = pyield(model, solution, strain_design.targets)
target_flux = solution.fluxes[pyield.product]
biomass = solution.fluxes[bpcy.biomass]
except SolveError:
_bpcy, _pyield, target_flux, biomass = np.nan, np.nan, np.nan, np.nan
return _bpcy, _pyield, target_flux, biomass
def fba(self,
measured_metabolites,
filter_by_subsystem=False,
add_constraints=False):
if add_constraints:
self.increasing_metabolite_constrains(measured_metabolites)
if measured_metabolites is not None:
self.set_objective_coefficients(measured_metabolites)
reactions = None
if filter_by_subsystem:
reactions = self.filter_reaction_by_subsystems()
return fba(self, reactions=reactions)
def evaluate_design(model, strain_design, pathway, aerobic, bpcy, pyield):
with model:
if not aerobic and 'EX_o2_e' in model.reactions:
model.reactions.EX_o2_e.lower_bound = 0
pathway.apply(model)
strain_design.apply(model)
try:
solution = fba(model, objective=model.biomass)
_bpcy = bpcy(model, solution, strain_design.targets)
_pyield = pyield(model, solution, strain_design.targets)
target_flux = solution.fluxes[pyield.product]
biomass = solution.fluxes[bpcy.biomass]
except OptimizationError:
_bpcy, _pyield, target_flux, biomass = np.nan, np.nan, np.nan, np.nan
return _bpcy, _pyield, target_flux, biomass
def evaluate_design(model, strain_design, pathway, aerobic, bpcy, pyield):
with model:
if not aerobic and 'EX_o2_e' in model.reactions:
model.reactions.EX_o2_e.lower_bound = 0
pathway.apply(model)
strain_design.apply(model)
try:
solution = fba(model, objective=model.biomass)
_bpcy = bpcy(model, solution, strain_design.targets)
_pyield = pyield(model, solution, strain_design.targets)
target_flux = solution.fluxes[pyield.product]
biomass = solution.fluxes[bpcy.biomass]
except OptimizationError:
_bpcy, _pyield, target_flux, biomass = np.nan, np.nan, np.nan, np.nan
return _bpcy, _pyield, target_flux, biomass
def test_apply_designs(self, model, diff_fva):
result = diff_fva.run()
works = []
for strain_design in result:
with model:
strain_design.apply(model)
try:
solution = fba(
model,
objective="Biomass_Ecoli_core_N_lp_w_fsh_GAM_rp__Nmet2"
)
works.append(solution["EX_succ_lp_e_rp_"] > 1e-6
and solution.objective_value > 1e-6)
except Infeasible:
works.append(False)
assert any(works)
def test_apply_designs(self, model):
result = DifferentialFVA(model,
model.reactions.EX_succ_lp_e_rp_,
points=5).run()
works = []
for strain_design in result:
with TimeMachine() as tm:
strain_design.apply(model, tm)
try:
solution = fba(
model,
objective="Biomass_Ecoli_core_N_lp_w_fsh_GAM_rp__Nmet2"
)
works.append(solution["EX_succ_lp_e_rp_"] > 1e-6
and solution.objective_value > 1e-6)
except Infeasible:
works.append(False)
assert any(works)
def _fba(model, analysis):
'''!
Private function - to simulate a model using Flux Balance
Analysis (FBA) or FBA-related methods, with Cameo.
@model String: Model acceptable by Cameo (see
http://cameo.bio/02-import-models.html).
@analysis String: Type of FBA to perform. Allowable types are
FBA (standard flux balance analysis) and pFBA (parsimonious
FBA).
'''
import cameo
if analysis == 'FBA':
print('Run flux balance analysis on model %s' % str(model))
return cameo.fba(model)
elif analysis == 'pFBA':
print('Run parsimonious flux balance analysis on model %s' \
% str(model))
return cameo.pfba(model)
def gene_deletion(self) -> pd.DataFrame:
model = self.read_model()
gene_status = []
gene_values = []
knockout_reactions = self.knockout_reactions_for_genes(self.model_path)
for gene in model.genes:
reaction_bounds = dict()
# knockout all reactions affected by gene by setting bounds zero
for rid in knockout_reactions[gene.id]:
reaction = model.reactions.get_by_id(rid)
reaction_bounds[reaction.id] = (
reaction.lower_bound,
reaction.upper_bound,
)
reaction.bounds = (0, 0)
try:
# run fba
result = fba(model)
value = result.objective_value
status = CuratorConstants.STATUS_OPTIMAL
except Exception as e:
logger.error(f"{e}")
value = CuratorConstants.VALUE_INFEASIBLE
status = CuratorConstants.STATUS_INFEASIBLE
gene_status.append(status)
gene_values.append(value)
# restore bounds
for rid, bounds in reaction_bounds.items():
model.reactions.get_by_id(rid).bounds = bounds[:]
return pd.DataFrame(
{
"model": self.model_path.name,
"objective": self.objective_id,
"gene": [gene.id for gene in model.genes],
"status": gene_status,
"value": gene_values,
}
)
def objective(self) -> pd.DataFrame:
model = self.read_model()
try:
# fbc optimization
result = fba(model)
value = result.objective_value
status = CuratorConstants.STATUS_OPTIMAL
except Exception as e:
logger.error(f"{e}")
value = CuratorConstants.VALUE_INFEASIBLE
status = CuratorConstants.STATUS_INFEASIBLE
return pd.DataFrame(
{
"model": self.model_path.name,
"objective": [self.objective_id],
"status": [status],
"value": [value],
}
)
def calculate_yield(model, source, product):
try:
flux_dist = fba(model, objective=product)
return flux_dist[product.id] / abs(flux_dist[source.id])
except OptimizationError:
return 0.0
def gimme(model,
expression_profile=None,
cutoff=None,
objective=None,
objective_dist=None,
fraction_of_optimum=0.9,
normalization=or2min_and2max,
condition=None,
not_measured_value=None,
*args,
**kwargs):
"""
Gene Inactivity Moderated by Metabolism and Expression (GIMME)[1]
Parameters
----------
model: SolverBased Model
A constraint based model
expression_profile: ExpressionProfile
An expression profile
cutoff: float
inactivity threshold
objective: str or other cameo compatible objective
The Minimal Required Functionalities (MRF)
objective_dist: FluxDistributionResult
A predetermined flux distribution for the objective can be provided (optional)
fraction_of_optimum: float
The fraction of the MRF
normalization: function
expression profile normalization function
condition: str, int or None
The condition from the expression profile. If None (default), the first condition on the profile will be used.
normalization: function
The normalization function to convert the gene expression profile into reaction expression profile (default: max)
Returns
-------
GimmeResult
References
----------
.. [1] Becker, S. a and Palsson, B. O. (2008). Context-specific metabolic networks are consistent with experiments.
PLoS Computational Biology, 4(5), e1000082. doi:10.1371/journal.pcbi.1000082
"""
assert isinstance(model, SolverBasedModel)
assert isinstance(expression_profile, ExpressionProfile)
assert isinstance(fraction_of_optimum, numbers.Number)
assert isinstance(cutoff, numbers.Number)
objective = model.objective if objective is None else objective
objective_dist = fba(
model,
objective=objective) if objective_dist is None else objective_dist
assert isinstance(objective_dist, FluxDistributionResult)
if objective.direction == 'max':
fix_obj_constraint = model.solver.interface.Constraint(
model.objective.expression,
lb=fraction_of_optimum * objective_dist.objective_value,
name="required metabolic functionalities")
else:
fix_obj_constraint = model.solver.interface.Constraint(
model.objective.expression,
ub=fraction_of_optimum * objective_dist.objective_value,
name="required metabolic functionalities")
objective_terms = list()
condition = expression_profile.conditions[
0] if condition is None else condition
not_measured_value = cutoff if not_measured_value is None else not_measured_value
reaction_profile = expression_profile.to_reaction_dict(
condition, model, not_measured_value, normalization)
coefficients = {
r: cutoff - exp if cutoff > exp else 0
for r, exp in six.iteritems(reaction_profile)
}
for rid, coefficient in six.iteritems(coefficients):
reaction = model.reactions.get_by_id(rid)
if coefficient > 0:
objective_terms.append(
coefficient *
(reaction.forward_variable + reaction.reverse_variable))
with TimeMachine() as tm:
gimme_objective = model.solver.interface.Objective(
Add(*objective_terms), direction="min")
tm(do=partial(setattr, model, "objective", gimme_objective),
undo=partial(setattr, model, "objective",
model.objective.expression))
tm(do=partial(model.solver.add, fix_obj_constraint),
undo=partial(model.solver.remove, [fix_obj_constraint]))
solution = model.solve()
return GimmeResult(solution.fluxes, solution.f, objective_dist.fluxes,
reaction_profile, cutoff)
def run(self,
product=None,
max_predictions=float("inf"),
min_production=.1,
timeout=None,
callback=None,
silent=False,
allow_native_exchanges=False):
"""Run pathway prediction for a desired product.
Parameters
----------
product : Metabolite, str
Metabolite or id or name of metabolite to find production pathways for.
max_predictions : int, optional
The maximum number of predictions to compute.
min_production : float
The minimum acceptable production flux to product.
timeout : int
The time limit [seconds] per attempted prediction.
callback : function
A function that takes a successfully predicted pathway.
silent : bool
If True will print the pathways and max flux values.
allow_native_exchanges: bool
If True, exchange reactions for native metabolites will be allowed.
Returns
-------
PathwayPredictions
The predicted pathways.
"""
product = self._find_product(product)
pathways = list()
with TimeMachine() as tm, self.model:
tm(do=partial(setattr, self.model.solver.configuration, 'timeout',
timeout),
undo=partial(setattr, self.model.solver.configuration,
'timeout',
self.model.solver.configuration.timeout))
try:
product_reaction = self.model.reactions.get_by_id('DM_' +
product.id)
except KeyError:
product_reaction = self.model.add_boundary(product,
type='demand')
product_reaction.lower_bound = min_production
counter = 1
while counter <= max_predictions:
logger.debug('Predicting pathway No. %d' % counter)
try:
self.model.slim_optimize(error_value=None)
except OptimizationError as e:
logger.error(
'No pathway could be predicted. Terminating pathway predictions.'
)
logger.error(e)
break
vars_to_cut = list()
for i, y_var_id in enumerate(self._y_vars_ids):
y_var = self.model.solver.variables[y_var_id]
if y_var.primal == 1.0:
vars_to_cut.append(y_var)
logger.info(vars_to_cut)
if len(vars_to_cut) == 0:
# no pathway found:
logger.info(
"It seems %s is a native product in model %s. "
"Let's see if we can find better heterologous pathways."
% (product, self.model))
# knockout adapter with native product
for adapter in self.adpater_reactions:
if product in adapter.metabolites:
logger.info(
'Knocking out adapter reaction %s containing native product.'
% adapter)
adapter.knock_out()
continue
pathway = [
self.model.reactions.get_by_id(y_var.name[2:])
for y_var in vars_to_cut
]
pathway_metabolites = set([
m for pathway_reaction in pathway
for m in pathway_reaction.metabolites
])
logger.info('Pathway predicted: %s' % '\t'.join([
r.build_reaction_string(use_metabolite_names=True)
for r in pathway
]))
pathway_metabolites.add(product)
# Figure out adapter reactions to include
adapters = [
adapter for adapter in self.adpater_reactions
if adapter.products[0] in pathway_metabolites
]
# Figure out exchange reactions to include
exchanges = [
exchange for exchange in self._exchanges
if abs(exchange.flux) > non_zero_flux_threshold
and exchange.id != product_reaction.id
]
if allow_native_exchanges:
exchanges = [
exchange for exchange in exchanges
if list(exchange.metabolites)[0] in pathway_metabolites
]
pathway = PathwayResult(pathway, exchanges, adapters,
product_reaction)
if not silent:
util.display_pathway(pathway, counter)
integer_cut = self.model.solver.interface.Constraint(
Add(*vars_to_cut),
name="integer_cut_" + str(counter),
ub=len(vars_to_cut) - 1)
logger.debug('Adding integer cut.')
tm(do=partial(self.model.solver.add, integer_cut),
undo=partial(self.model.solver.remove, integer_cut))
# Test pathway in the original model
with self.original_model:
pathway.apply(self.original_model)
try:
solution = fba(self.original_model,
objective=pathway.product.id)
except OptimizationError as e:
logger.error(e)
logger.error(
"Addition of pathway {} made the model unsolvable. "
"Skipping pathway.".format(pathway))
continue
else:
if solution[
pathway.product.id] > non_zero_flux_threshold:
pathways.append(pathway)
if not silent:
print("Max flux: %.5f" %
solution[pathway.product.id])
else:
logger.error(
"Pathway {} couldn't be verified. Production flux {}"
"is below requirement {}. Skipping pathway.".
format(pathway, solution[pathway.product.id],
non_zero_flux_threshold))
finally:
counter += 1
if callback is not None:
callback(pathway)
return PathwayPredictions(pathways)
def gimme(
model,
expression_profile=None,
cutoff=None,
objective=None,
objective_dist=None,
fraction_of_optimum=0.9,
normalization=or2min_and2max,
condition=None,
not_measured_value=None,
*args,
**kwargs
):
"""
Gene Inactivity Moderated by Metabolism and Expression (GIMME)[1]
Parameters
----------
model: SolverBased Model
A constraint based model
expression_profile: ExpressionProfile
An expression profile
cutoff: float
inactivity threshold
objective: str or other cameo compatible objective
The Minimal Required Functionalities (MRF)
objective_dist: FluxDistributionResult
A predetermined flux distribution for the objective can be provided (optional)
fraction_of_optimum: float
The fraction of the MRF
normalization: function
expression profile normalization function
condition: str, int or None
The condition from the expression profile. If None (default), the first condition on the profile will be used.
normalization: function
The normalization function to convert the gene expression profile into reaction expression profile (default: max)
Returns
-------
GimmeResult
References
----------
.. [1] Becker, S. a and Palsson, B. O. (2008). Context-specific metabolic networks are consistent with experiments.
PLoS Computational Biology, 4(5), e1000082. doi:10.1371/journal.pcbi.1000082
"""
assert isinstance(model, SolverBasedModel)
assert isinstance(expression_profile, ExpressionProfile)
assert isinstance(fraction_of_optimum, numbers.Number)
assert isinstance(cutoff, numbers.Number)
objective = model.objective if objective is None else objective
objective_dist = fba(model, objective) if objective_dist is None else objective_dist
assert isinstance(objective_dist, FluxDistributionResult)
if objective.direction == "max":
fix_obj_constraint = model.solver.interface.Constraint(
model.objective.expression,
lb=fraction_of_optimum * objective_dist.objective_value,
name="required metabolic functionalities",
)
else:
fix_obj_constraint = model.solver.interface.Constraint(
model.objective.expression,
ub=fraction_of_optimum * objective_dist.objective_value,
name="required metabolic functionalities",
)
objective_terms = list()
condition = expression_profile.conditions[0] if condition is None else condition
not_measured_value = cutoff if not_measured_value is None else not_measured_value
reaction_profile = expression_profile.to_reaction_dict(condition, model, not_measured_value, normalization)
coefficients = {r: cutoff - exp if cutoff > exp else 0 for r, exp in six.iteritems(reaction_profile)}
for rid, coefficient in six.iteritems(coefficients):
reaction = model.reactions.get_by_id(rid)
if coefficient > 0:
objective_terms.append(coefficient * (reaction.forward_variable + reaction.reverse_variable))
with TimeMachine() as tm:
gimme_objective = model.solver.interface.Objective(Add(*objective_terms), direction="min")
tm(
do=partial(setattr, model, "objective", gimme_objective),
undo=partial(setattr, model, "objective", model.objective.expression),
)
tm(
do=partial(model.solver._add_constraint, fix_obj_constraint),
undo=partial(model.solver._remove_constraints, [fix_obj_constraint]),
)
solution = model.solve()
return GimmeResult(solution.fluxes, solution.f, objective_dist.fluxes, reaction_profile, cutoff)
def calculate_yield(model, source, product):
try:
flux_dist = fba(model, objective=product)
return flux_dist[product.id] / abs(flux_dist[source.id])
except OptimizationError:
return 0.0
'''
Created on 4 Dec 2018
@author: mate
load in CD4 model, extract all proteins and save them to a new .csv
'''
from cameo import load_model, fba
model = load_model('models/CD4T1670.xml')
print(model.reactions)
new_solution = model.optimize()
fba_result = fba(model)
print(fba_result)
print(fba_result.data_frame)
print(len(fba_result.data_frame))
# print(model.reactions)
for reactionO in model.reactions:
for reactionI in reactionO.genes:
print(reactionI)
fba_result.display_on_map("RECON1.Carbohydrate metabolism")
