def run_detfl(yaml_file, uptake_funs, medium_funs=dict(), model=None, ini_sol=None): # pass in
params = read_config(yaml_file)
for key, value in params.items():
print('%s: %s' % (key, value))
if model is None:
model = load_json_model(params['model'])
standard_solver_config(model)
model.solver.configuration.verbosity = params['options']['verbose']
time_data = run_dynamic_etfl(model,
timestep=params['simulation']['timestep'],
tfinal=params['simulation']['tfinal'],
uptake_fun=uptake_funs,
medium_fun=medium_funs,
uptake_enz=params['assumptions']['uptake_enz'],
S0=params['assumptions']['S0'],
X0=params['assumptions']['X0'],
inplace=params['options']['inplace'],
initial_solution = ini_sol,
chebyshev_include = params['options']['chebyshev_include'],
dynamic_constraints = params['options']['constraints']
)
# plot_dynamics(model, time_data)
out_path = join('outputs',params['tag']+get_timestr())
makedirs(out_path)
time_data.to_csv(join(out_path,'solution.csv'))
write_yaml(params,join(out_path,'config.yaml'))
return time_data
def make_thermo_model():
vanilla_model = get_model(solver)
name = 'iJO1366_T1E0N0_{}'.format(get_timestr())
vanilla_model.reactions.EX_glc__D_e.lower_bound = -1 * glc_uptake - glc_uptake_std
vanilla_model.reactions.EX_glc__D_e.upper_bound = -1 * glc_uptake + glc_uptake_std
vanilla_model.objective = growth_reaction_id
vanilla_model.optimize()
thermo_data, lexicon, compartment_data = get_thermo_data()
ecoli = ThermoMEModel(
thermo_data,
model=vanilla_model,
name=name,
)
need_relax = False
try:
ecoli.optimize()
except AttributeError:
need_relax = True
if need_relax:
final_model, slack_model, relax_table = relax_dgo(ecoli)
else:
final_model = ecoli
from pytfa.io.json import save_json_model
save_json_model(final_model, 'models/' + name)
def make_fba_model():
ecoli = get_model(solver)
ecoli.reactions.EX_glc__D_e.lower_bound = -1 * glc_uptake - glc_uptake_std
ecoli.reactions.EX_glc__D_e.upper_bound = -1 * glc_uptake + glc_uptake_std
# ecoli.objective = growth_reaction_id
ecoli.optimize()
from cobra.io.json import save_json_model
save_json_model(ecoli,
'models/iJO1366_T0E0N0_{}.json'.format(get_timestr()))
def create_model(has_thermo,
has_expression,
has_allocation,
kcat_mode='kmax',
infer_missing_enz=False,
additional_enz=None,
free_rib_ratio=0.2,
free_rnap_ratio=0.75,
add_displacement=False,
n_mu_bins=128,
name_suffix='',
kcat_overrides=None):
#------------------------------------------------------------
# Initialisation
#------------------------------------------------------------
assert has_expression == True
# this hack works because we are using the solver switch to update the var
# names in the solver but really we should not do this
# TODO: clean up model.sanitize_varnames
vanilla_model = get_model('optlang-glpk')
vanilla_model.reactions.EX_glc__D_e.lower_bound = -1 * glc_uptake - glc_uptake_std
vanilla_model.reactions.EX_glc__D_e.upper_bound = -1 * glc_uptake + glc_uptake_std
vanilla_model.objective = growth_reaction_id
fba_sol = vanilla_model.slim_optimize()
# vanilla_model.reactions.get_by_id(growth_reaction_id).lower_bound = observed_growth
# fva = flux_variability_analysis(vanilla_model)
# vanilla_model.reactions.get_by_id(growth_reaction_id).lower_bound = 0
# original_bounds = pd.DataFrame.from_dict(
# {r.id:(r.lower_bound, r.upper_bound)
# for r in vanilla_model.reactions}, orient = 'index')
# original_bounds.columns = ['lb','ub']
mu_0 = fba_sol
mu_range = [0, 3.5]
n_mu_bins = n_mu_bins
time_str = get_timestr()
coupling_dict = get_coupling_dict(vanilla_model,
mode=kcat_mode,
atps_name='ATPS4rpp',
infer_missing_enz=infer_missing_enz)
if additional_enz is not None:
additional_dict = get_transporters_coupling(
model=vanilla_model, additional_enz=additional_enz)
additional_dict.update(coupling_dict)
coupling_dict = additional_dict
if kcat_overrides is not None:
for rxn, enz_list in coupling_dict.items():
for e in enz_list:
if e.id in kcat_overrides:
prev_kcat = e.kcat_fwd
new_kcat = kcat_overrides[e.id]
e.kcat_fwd = new_kcat
e.kcat_bwd = new_kcat
print('Replaced kcat for {}: {} <-- {} s-1'.format(
e.id, prev_kcat / 3600, new_kcat / 3600))
aa_dict, rna_nucleotides, rna_nucleotides_mp, dna_nucleotides = get_monomers_dict(
)
essentials = get_essentials()
# Initialize the model
model_name = 'ETFL' if has_thermo else 'EFL'
model_name = ('v' + model_name) if has_allocation else model_name
model_name = (model_name +
'_{}'.format(name_suffix)) if name_suffix else model_name
model_name = (model_name +
'_infer') if bool(infer_missing_enz) else model_name
name = 'iJO1366_{}_{}_enz_{}_bins_{}.json'.format(model_name,
len(coupling_dict),
n_mu_bins, time_str)
if has_thermo:
thermo_data, lexicon, compartment_data = get_thermo_data()
ecoli = ThermoMEModel(
thermo_data,
model=vanilla_model,
growth_reaction=growth_reaction_id,
mu_range=mu_range,
n_mu_bins=n_mu_bins,
name=name,
)
else:
ecoli = MEModel(
model=vanilla_model,
growth_reaction=growth_reaction_id,
mu_range=mu_range,
n_mu_bins=n_mu_bins,
name=name,
)
ecoli.name = name
ecoli.logger.setLevel(logging.WARNING)
ecoli.sloppy = True
# apply_bounds(ecoli,fva)
ecoli.solver = solver
standard_solver_config(ecoli, verbose=False)
if has_thermo:
# Annotate the cobra_model
annotate_from_lexicon(ecoli, lexicon)
apply_compartment_data(ecoli, compartment_data)
# TFA conversion
ecoli.prepare()
ecoli.convert(add_displacement=add_displacement)
mrna_dict = get_mrna_dict(ecoli)
nt_sequences = get_nt_sequences()
rnap = get_rnap()
# rnap.kcat_fwd *= 0.5
rib = get_rib()
# Remove nucleotides and amino acids from biomass reaction as they will be
# taken into account by the expression
remove_from_biomass_equation(model=ecoli,
nt_dict=rna_nucleotides,
aa_dict=aa_dict,
essentials_dict=essentials)
##########################
## MODEL CREATION ##
##########################
ecoli.add_nucleotide_sequences(nt_sequences)
ecoli.add_essentials(essentials=essentials,
aa_dict=aa_dict,
rna_nucleotides=rna_nucleotides,
rna_nucleotides_mp=rna_nucleotides_mp)
ecoli.add_mrnas(mrna_dict.values())
ecoli.add_ribosome(rib, free_rib_ratio)
ecoli.add_rnap(rnap, free_rnap_ratio)
ecoli.build_expression()
ecoli.add_enzymatic_coupling(coupling_dict)
if has_allocation:
nt_ratios, aa_ratios = get_ratios()
chromosome_len, gc_ratio = get_ecoli_gen_stats()
kdeg_mrna, mrna_length_avg = get_mrna_metrics()
kdeg_enz, peptide_length_avg = get_enz_metrics()
neidhardt_mu, neidhardt_rrel, neidhardt_prel, neidhardt_drel = get_neidhardt_data(
)
ecoli.add_dummies(nt_ratios=nt_ratios,
mrna_kdeg=kdeg_mrna,
mrna_length=mrna_length_avg,
aa_ratios=aa_ratios,
enzyme_kdeg=kdeg_enz,
peptide_length=peptide_length_avg)
add_protein_mass_requirement(ecoli, neidhardt_mu, neidhardt_prel)
add_rna_mass_requirement(ecoli, neidhardt_mu, neidhardt_rrel)
add_dna_mass_requirement(ecoli,
mu_values=neidhardt_mu,
dna_rel=neidhardt_drel,
gc_ratio=gc_ratio,
chromosome_len=chromosome_len,
dna_dict=dna_nucleotides)
dna_pol = get_dna_polymerase()
ecoli.add_enzymatic_coupling({'DNA_formation': [
dna_pol,
]})
# Need to put after, because dummy has to be taken into account if used.
ecoli.populate_expression()
ecoli.add_trna_mass_balances()
ecoli.print_info()
ecoli.growth_reaction.lower_bound = observed_growth - 1 * observed_growth_std
need_relax = False
ecoli.repair()
try:
ecoli.optimize()
except (AttributeError, SolverError):
need_relax = True
if has_thermo and need_relax:
# final_model, slack_model, relax_table = relax_dgo(ecoli)
final_model, slack_model, relax_table = relax_dgo(ecoli, in_place=True)
else:
final_model = ecoli
final_model.growth_reaction.lower_bound = 0
# apply_bounds(ecoli, original_bounds)
solution = final_model.optimize()
print_standard_sol(final_model)
filepath = 'models/{}'.format(final_model.name)
save_json_model(final_model, filepath)
final_model.logger.info('Build complete for model {}'.format(
final_model.name))
return final_model
def create_model(has_thermo, has_expression, has_neidhardt, n_mu_bins=128):
#------------------------------------------------------------
# Initialisation
#------------------------------------------------------------
assert has_expression == True
# this hack works because we are using the solver switch to update the var
# names in the solver but really we should not do this
# TODO: clean up model.sanitize_varnames
vanilla_model = get_model('optlang-glpk')
vanilla_model.reactions.EX_glc__D_e.lower_bound = -1 * glc_uptake - glc_uptake_std
vanilla_model.reactions.EX_glc__D_e.upper_bound = -1 * glc_uptake + glc_uptake_std
vanilla_model.objective = growth_reaction_id
fba_sol = vanilla_model.slim_optimize()
mu_0 = fba_sol
mu_range = [0, 3.5]
n_mu_bins = n_mu_bins
time_str = get_timestr()
coupling_dict = get_coupling_dict(
vanilla_model,
mode='kmax',
# mode = 'kcat',
atps_name='ATPS4rpp',
infer_missing_enz=True)
# coupling_dict = get_lloyd_coupling_dict(vanilla_model)
aa_dict, rna_nucleotides, rna_nucleotides_mp, dna_nucleotides = get_monomers_dict(
)
essentials = get_essentials()
# Initialize the model
name = 'iJO1366_T{:1}E{:1}N{:1}_{}_enz_{}_bins_{}.json'.format(
has_thermo, has_expression, has_neidhardt, len(coupling_dict),
n_mu_bins, time_str)
if has_thermo:
thermo_data, lexicon, compartment_data = get_thermo_data()
ecoli = ThermoMEModel(
thermo_data,
model=vanilla_model,
growth_reaction=growth_reaction_id,
mu_range=mu_range,
n_mu_bins=n_mu_bins,
name=name,
)
else:
ecoli = MEModel(
model=vanilla_model,
growth_reaction=growth_reaction_id,
mu_range=mu_range,
n_mu_bins=n_mu_bins,
name=name,
)
ecoli.name = name
ecoli.logger.setLevel(logging.WARNING)
ecoli.solver = solver
standard_solver_config(ecoli)
if has_thermo:
# Annotate the cobra_model
annotate_from_lexicon(ecoli, lexicon)
apply_compartment_data(ecoli, compartment_data)
# TFA conversion
ecoli.prepare()
# ecoli.reactions.GLUDy.thermo['computed'] = False
# ecoli.reactions.DHAtpp.thermo['computed'] = False
# ecoli.reactions.MLTP2.thermo['computed'] = False
# ecoli.reactions.G3PD2.thermo['computed'] = False
ecoli.reactions.MECDPS.thermo['computed'] = False
ecoli.reactions.NDPK4.thermo['computed'] = False
ecoli.reactions.TMDPP.thermo['computed'] = False
ecoli.reactions.ARGAGMt7pp.thermo['computed'] = False
ecoli.convert() #add_displacement = True)
mrna_dict = get_mrna_dict(ecoli)
nt_sequences = get_nt_sequences()
rnap = get_rnap()
rib = get_rib()
# Remove nucleotides and amino acids from biomass reaction as they will be
# taken into account by the expression
remove_from_biomass_equation(
model=ecoli,
nt_dict=rna_nucleotides,
aa_dict=aa_dict,
atp_id=essentials['atp'],
adp_id=essentials['adp'],
pi_id=essentials['pi'],
h2o_id=essentials['h2o'],
h_id=essentials['h'],
)
##########################
## MODEL CREATION ##
##########################
ecoli.add_nucleotide_sequences(nt_sequences)
ecoli.add_essentials(essentials=essentials,
aa_dict=aa_dict,
rna_nucleotides=rna_nucleotides,
rna_nucleotides_mp=rna_nucleotides_mp)
ecoli.add_mrnas(mrna_dict.values())
ecoli.add_ribosome(rib, free_ratio=0.2)
# http://bionumbers.hms.harvard.edu/bionumber.aspx?id=102348&ver=1&trm=rna%20polymerase%20half%20life&org=
# Name Fraction of active RNA Polymerase
# Bionumber ID 102348
# Value 0.17-0.3 unitless
# Source Bremer, H., Dennis, P. P. (1996) Modulation of chemical composition and other parameters of the cell by growth rate.
# Neidhardt, et al. eds. Escherichia coli and Salmonella typhimurium: Cellular
# and Molecular Biology, 2nd ed. chapter 97 Table 1
ecoli.add_rnap(rnap, free_ratio=0.75)
ecoli.build_expression()
ecoli.add_enzymatic_coupling(coupling_dict)
if has_neidhardt:
nt_ratios, aa_ratios = get_ratios()
chromosome_len, gc_ratio = get_ecoli_gen_stats()
kdeg_mrna, mrna_length_avg = get_mrna_metrics()
kdeg_enz, peptide_length_avg = get_enz_metrics()
neidhardt_mu, neidhardt_rrel, neidhardt_prel, neidhardt_drel = get_neidhardt_data(
)
ecoli.add_interpolation_variables()
ecoli.add_dummies(nt_ratios=nt_ratios,
mrna_kdeg=kdeg_mrna,
mrna_length=mrna_length_avg,
aa_ratios=aa_ratios,
enzyme_kdeg=kdeg_enz,
peptide_length=peptide_length_avg)
ecoli.add_protein_mass_requirement(neidhardt_mu, neidhardt_prel)
ecoli.add_rna_mass_requirement(neidhardt_mu, neidhardt_rrel)
ecoli.add_dna_mass_requirement(mu_values=neidhardt_mu,
dna_rel=neidhardt_drel,
gc_ratio=gc_ratio,
chromosome_len=chromosome_len,
dna_dict=dna_nucleotides)
# Need to put after, because dummy has to be taken into account if used.
ecoli.populate_expression()
ecoli.add_trna_mass_balances()
ecoli.print_info()
ecoli.growth_reaction.lower_bound = observed_growth - 3 * observed_growth_std
need_relax = False
ecoli.repair()
try:
ecoli.optimize()
except (AttributeError, SolverError):
need_relax = True
# from ipdb import set_trace; set_trace()
if has_thermo and need_relax:
final_model, slack_model, relax_table = relax_dgo(ecoli)
# final_model, slack_model, relax_table = relax_dgo(ecoli, in_place = True)
else:
final_model = ecoli
final_model.growth_reaction.lower_bound = 0
solution = final_model.optimize()
print('Objective : {}'.format(
final_model.solution.objective_value))
print(' - Glucose uptake : {}'.format(
final_model.reactions.EX_glc__D_e.flux))
print(' - Growth : {}'.format(final_model.growth_reaction.flux))
print(' - Ribosomes produced: {}'.format(final_model.ribosome.X))
print(' - RNAP produced: {}'.format(final_model.rnap.X))
try:
print(' - DNA produced: {}'.format(final_model.solution.raw.DN_DNA))
except AttributeError:
pass
filepath = 'models/{}'.format(final_model.name)
# save_json_model(final_model, filepath)
final_model.logger.info('Build complete for model {}'.format(
final_model.name))
return final_model
def create_tfa_model(add_displacement=False):
#------------------------------------------------------------
# Initialisation
#------------------------------------------------------------
time_str = get_timestr()
name = 'iJO1366_TFA_{}.json'.format(time_str)
# this hack works because we are using the solver switch to update the var
# names in the solver but really we should not do this
# TODO: clean up model.sanitize_varnames
vanilla_model = get_model('optlang-glpk')
vanilla_model.reactions.EX_glc__D_e.lower_bound = -1 * glc_uptake - glc_uptake_std
vanilla_model.reactions.EX_glc__D_e.upper_bound = -1 * glc_uptake + glc_uptake_std
vanilla_model.objective = growth_reaction_id
fba_sol = vanilla_model.slim_optimize()
thermo_data, lexicon, compartment_data = get_thermo_data()
ecoli = ThermoModel(
thermo_data=thermo_data,
model=vanilla_model,
name=name,
)
ecoli.name = name
ecoli.logger.setLevel(logging.WARNING)
ecoli.sloppy = True
# apply_bounds(ecoli,fva)
ecoli.solver = solver
annotate_from_lexicon(ecoli, lexicon)
apply_compartment_data(ecoli, compartment_data)
# TFA conversion
ecoli.prepare()
ecoli.convert(add_displacement=add_displacement)
ecoli.print_info()
ecoli.reactions.get_by_id(growth_reaction_id).lower_bound = observed_growth - \
1*observed_growth_std
need_relax = False
ecoli.repair()
try:
ecoli.optimize()
except (AttributeError, SolverError):
need_relax = True
if need_relax:
# final_model, slack_model, relax_table = relax_dgo(ecoli)
final_model, slack_model, relax_table = relax_dgo(ecoli, in_place=True)
else:
final_model = ecoli
final_model.reactions.get_by_id(growth_reaction_id).lower_bound = 0
# apply_bounds(ecoli, original_bounds)
solution = final_model.optimize()
print('Objective : {}'.format(
final_model.solution.objective_value))
print(' - Glucose uptake : {}'.format(
final_model.reactions.EX_glc__D_e.flux))
filepath = 'models/{}'.format(final_model.name)
save_json_model(final_model, filepath)
final_model.logger.info('Build complete for model {}'.format(
final_model.name))
return final_model