def test_error_model(pheno_path):
model_prop = Model(pheno_path)
assert model_prop.error_model == 'PROP'
model_add_str = re.sub(
r'Y=F\+W\*EPS\(1\)',
'Y=F+EPS(1)',
str(model_prop),
)
model_add = Model(StringIO(model_add_str))
assert model_add.error_model == 'ADD'
model_add_prop_str = re.sub(
r'Y=F\+W\*EPS\(1\)',
'Y=EPS(1)*F+EPS(2)+F',
str(model_prop),
)
model_add_prop = Model(StringIO(model_add_prop_str))
assert model_add_prop.error_model == 'ADD_PROP'
model_none_str = re.sub(
r'Y=F\+W\*EPS\(1\)',
'Y=F',
str(model_prop),
)
model_none = Model(StringIO(model_none_str))
assert model_none.error_model == 'NONE'
def test_copy(datadir):
path = datadir / 'minimal.mod'
model = Model(path)
copy = model.copy()
assert id(model) != id(copy)
assert model.statements[0].expression == \
Symbol('THETA(1)', real=True) + Symbol('ETA(1)', real=True) + Symbol('ERR(1)', real=True)
def test_combined_error_model(testdata):
model = Model(testdata / 'nonmem' / 'pheno.mod')
combined_error(model)
model.update_source()
assert str(model).split('\n')[11] == 'Y = EPS(1)*F + EPS(2) + F'
assert str(model).split('\n')[17] == '$SIGMA 0.09 ; sigma_prop'
assert str(model).split('\n')[18] == '$SIGMA 11.2225 ; sigma_add'
def test_get_etas(pheno_path, testdata):
model = Model(pheno_path)
etas = _get_etas(model, ['ETA(1)'])
assert len(etas) == 1
etas = _get_etas(model, ['ETA(1)', 'CL'], include_symbols=True)
assert len(etas) == 1
etas = _get_etas(model, ['ETA(1)', 'V'], include_symbols=True)
assert len(etas) == 2
with pytest.raises(KeyError):
_get_etas(model, ['ETA(23)'])
model = Model(testdata / 'nonmem' / 'pheno_block.mod')
rvs = _get_etas(model, None)
assert rvs[0].name == 'ETA(1)'
model.parameters.fix = {'OMEGA(1,1)': True}
rvs = _get_etas(model, None)
assert rvs[0].name == 'ETA(2)'
model = Model(testdata / 'nonmem' / 'pheno_block.mod')
model.random_variables['ETA(1)'].level = 'IOV'
rvs = _get_etas(model, None)
assert rvs[0].name == 'ETA(2)'
def test_eq_modelstatements(testdata):
model_min = Model(testdata / 'nonmem' / 'minimal.mod')
model_pheno = Model(testdata / 'nonmem' / 'pheno_real.mod')
assert model_min.statements == model_min.statements
assert model_pheno.statements == model_pheno.statements
assert model_min.statements != model_pheno.statements
def test_michaelis_menten_elimination(testdata):
model = Model(testdata / 'nonmem' / 'pheno.mod')
michaelis_menten_elimination(model)
model.update_source()
correct = """$PROBLEM PHENOBARB SIMPLE MODEL
$DATA pheno.dta IGNORE=@
$INPUT ID TIME AMT WGT APGR DV
$SUBROUTINE ADVAN6 TOL=3
$MODEL COMPARTMENT=(CENTRAL DEFDOSE)
$PK
CLMM = THETA(3)
KM = THETA(2)
V = THETA(1)*EXP(ETA(1))
S1=V
$DES
DADT(1) = -A(1)*CLMM*KM/(V*(A(1)/V + KM))
$ERROR
Y=F+F*EPS(1)
$THETA (0,1.00916) ; TVV
$THETA (0,135.8) ; POP_KM
$THETA (0,0.00469307) ; POP_CLMM
$OMEGA 0.031128 ; IVV
$SIGMA 0.013241
$ESTIMATION METHOD=1 INTERACTION
"""
assert str(model) == correct
def test_iiv_on_ruv(pheno_path, epsilons, same_eta, err_ref, omega_ref):
model = Model(pheno_path)
model_str = str(model)
model_more_eps = re.sub('IPRED=F\nIRES=DV-IPRED',
'IPRED=F+EPS(2)\nIRES=DV-IPRED+EPS(3)', model_str)
model_sigma = re.sub(r'\$SIGMA 0.013241',
'$SIGMA 0.013241\n$SIGMA 0.1\n$SIGMA 0.1',
model_more_eps)
model.control_stream = NMTranParser().parse(model_sigma)
iiv_on_ruv(model, epsilons, same_eta)
model.update_source()
err_rec = model.control_stream.get_records('ERROR')[0]
assert str(
err_rec) == f'$ERROR\n' f'W=F\n' f'{err_ref}' f'IWRES=IRES/W\n\n'
omega_rec = ''.join(
str(rec) for rec in model.control_stream.get_records('OMEGA'))
assert omega_rec == (f'$OMEGA DIAGONAL(2)\n'
f' 0.0309626 ; IVCL\n'
f' 0.031128 ; IVV\n\n'
f'{omega_ref}\n')
def test_add_parameters(pheno_path, param_new, init_expected, buf_new):
model = Model(pheno_path)
pset = model.parameters
assert len(pset) == 6
pset.append(param_new)
model.parameters = pset
model.update_source()
assert len(pset) == 7
assert model.parameters[param_new.name].init == init_expected
parser = NMTranParser()
stream = parser.parse(str(model))
assert str(model.control_stream) == str(stream)
rec_ref = (
f'$THETA (0,0.00469307) ; PTVCL\n'
f'$THETA (0,1.00916) ; PTVV\n'
f'$THETA (-.99,.1)\n'
f'{buf_new}\n'
)
rec_mod = ''
for rec in model.control_stream.get_records('THETA'):
rec_mod += str(rec)
assert rec_ref == rec_mod
def test_parameter_estimates(pheno_path):
with ConfigurationContext(nonmem.conf, parameter_names=['basic']):
res = Model(pheno_path).modelfit_results
pe = res.parameter_estimates
assert len(pe) == 6
assert pe['THETA(1)'] == 4.69555e-3
assert pe['OMEGA(2,2)'] == 2.7906e-2
pe_sd = res.parameter_estimates_sdcorr
correct = pd.Series(
{
'THETA(1)': 0.00469555,
'THETA(2)': 0.984258,
'THETA(3)': 0.158920,
'OMEGA(1,1)': 0.171321,
'OMEGA(2,2)': 0.167051,
'SIGMA(1,1)': 0.115069,
}
)
correct.name = 'estimates'
pd.testing.assert_series_equal(pe_sd, correct)
with ConfigurationContext(nonmem.conf, parameter_names=['comment', 'basic']):
res = Model(pheno_path).modelfit_results
pe = res.parameter_estimates
assert len(pe) == 6
assert pe['PTVCL'] == 4.69555e-3
assert pe['IVV'] == 2.7906e-2
def test_proportional_error_model_log(testdata):
model = Model(testdata / 'nonmem' / 'pheno.mod')
model.statements[5] = Assignment('Y', 'F')
proportional_error(model, data_trans='log(Y)')
model.update_source()
assert str(model).split('\n')[11] == 'Y = LOG(F) + EPS(1)'
assert str(model).split('\n')[17] == '$SIGMA 0.09 ; sigma'
def test_remove_eta(pheno_path):
model = Model(pheno_path)
rvs = model.random_variables
eta1 = rvs['ETA(1)']
del rvs[eta1]
model.update_source()
assert str(model).split('\n')[12] == 'V = TVV*EXP(ETA(1))'
def psn_frem_results(path,
force_posdef_covmatrix=False,
force_posdef_samples=500,
method=None):
""" Create frem results from a PsN FREM run
:param path: Path to PsN frem run directory
:return: A :class:`FREMResults` object
"""
path = Path(path)
model_4_path = path / 'final_models' / 'model_4.mod'
if not model_4_path.is_file():
raise IOError(f'Could not find FREM model 4: {str(model_4_path)}')
model_4 = Model(model_4_path)
if model_4.modelfit_results is None:
raise ValueError('Model 4 has no results')
cov_model = None
if method == 'cov_sampling':
try:
model_4.modelfit_results.covariance_matrix
except Exception:
model_4b_path = path / 'final_models' / 'model_4b.mod'
try:
model_4b = Model(model_4b_path)
except FileNotFoundError:
pass
else:
cov_model = model_4b
with open(path / 'covariates_summary.csv') as covsum:
covsum.readline()
raw_cov_list = covsum.readline()
all_covariates = raw_cov_list[1:].rstrip().split(',')
nunique = model_4.dataset.pharmpy.baselines[all_covariates].nunique()
continuous = list(nunique.index[nunique != 2])
categorical = list(nunique.index[nunique == 2])
# FIXME: Not introducing yaml parser in pharmpy just yet. Options should be collected
# differently. Perhaps using json
with open(path / 'meta.yaml') as meta:
for row in meta:
row = row.strip()
if row.startswith('rescale: 1'):
rescale = True
elif row.startswith('rescale: 0'):
rescale = False
res = calculate_results(model_4,
continuous,
categorical,
method=method,
force_posdef_covmatrix=force_posdef_covmatrix,
force_posdef_samples=force_posdef_samples,
cov_model=cov_model,
rescale=rescale)
return res
def test_update_inits(pheno_path):
model = Model(pheno_path)
model.update_inits()
with ConfigurationContext(conf, parameter_names=['comment', 'basic']):
model = Model(pheno_path)
model.update_inits()
model.update_source()
def test_create_model3b(testdata):
model3 = Model(testdata / 'nonmem' / 'frem' / 'pheno' / 'model_3.mod')
model1b = Model(testdata / 'nonmem' / 'pheno_real.mod')
model3b = create_model3b(model1b, model3, 2)
pset = model3b.parameters
assert pset['OMEGA(3,1)'].init == approx(0.02560327)
assert pset['THETA(1)'].init == 0.00469555
assert model3b.name == 'model_3b'
def test_bootstrap(tmp_path, testdata):
with TemporaryDirectoryChanger(tmp_path):
shutil.copy2(testdata / 'nonmem' / 'pheno.mod', tmp_path)
shutil.copy2(testdata / 'nonmem' / 'pheno.dta', tmp_path)
model = Model('pheno.mod')
model.dataset_path = tmp_path / 'pheno.dta'
res = Bootstrap(model, 3).run()
assert len(res.parameter_estimates) == 3
def test_copy(datadir):
path = datadir / 'minimal.mod'
model = Model(path)
copy = model.copy()
assert id(model) != id(copy)
assert model.statements[0].expression == symbol('THETA(1)') + symbol('ETA(1)') + symbol(
'EPS(1)'
)
def test_additive_error_model(testdata):
model = Model(testdata / 'nonmem' / 'pheno.mod')
additive_error(model)
model.update_source()
assert str(model).split('\n')[11] == 'Y = F + EPS(1)'
assert str(model).split('\n')[17] == '$SIGMA 11.2225 ; sigma'
before = str(model)
additive_error(model) # One more time and nothing should change
assert before == str(model)
def test_find_depot(testdata):
model = Model(testdata / 'nonmem' / 'modeling' / 'pheno_advan2.mod')
assert model.statements.ode_system.find_depot(model.statements).name == 'DEPOT'
model = Model(testdata / 'nonmem' / 'modeling' / 'pheno_advan1.mod')
assert model.statements.ode_system.find_depot(model.statements) is None
model = Model(testdata / 'nonmem' / 'modeling' / 'pheno_advan5_depot.mod')
assert model.statements.ode_system.find_depot(model.statements).name == 'DEPOT'
model = Model(testdata / 'nonmem' / 'modeling' / 'pheno_advan5_nodepot.mod')
assert model.statements.ode_system.find_depot(model.statements) is None
def test_des(testdata, model_path, transformation):
model_ref = Model(testdata / model_path)
transformation(model_ref)
model_ref.update_source()
model_des = Model(StringIO(str(model_ref)))
model_des.source.path = model_ref.source.path # To be able to find dataset
assert model_ref.statements.ode_system == model_des.statements.ode_system
def test_fix_parameters(testdata):
model = Model(testdata / 'nonmem' / 'minimal.mod')
assert not model.parameters['THETA(1)'].fix
fix_parameters(model, ['THETA(1)'])
assert model.parameters['THETA(1)'].fix
model = Model(testdata / 'nonmem' / 'minimal.mod')
assert not model.parameters['THETA(1)'].fix
fix_parameters(model, 'THETA(1)')
assert model.parameters['THETA(1)'].fix
def test_fit_single(tmp_path, testdata):
with TemporaryDirectoryChanger(tmp_path):
shutil.copy2(testdata / 'nonmem' / 'pheno.mod', tmp_path)
shutil.copy2(testdata / 'nonmem' / 'pheno.dta', tmp_path)
model = Model('pheno.mod')
model.dataset_path = tmp_path / 'pheno.dta'
modeling.fit(model)
rundir = tmp_path / 'modelfit_dir1'
assert model.modelfit_results.ofv == pytest.approx(730.8947268137308)
assert rundir.is_dir()
def test_combined_error_model_log(testdata):
model = Model(testdata / 'nonmem' / 'pheno.mod')
combined_error(model, data_trans='log(Y)')
model.update_source()
assert str(model).split('\n')[11] == 'Y = LOG(F) + EPS(2)/F + EPS(1)'
assert str(model).split('\n')[17] == '$SIGMA 0.09 ; sigma_prop'
assert str(model).split('\n')[18] == '$SIGMA 11.2225 ; sigma_add'
before = str(model)
combined_error(model) # One more time and nothing should change
assert before == str(model)
def test_simeval(testdata):
orig = Model(testdata / 'nonmem' / 'pheno.mod')
base = Model(testdata / 'nonmem' / 'qa' / 'pheno_linbase.mod')
simeval_res = read_results(testdata / 'nonmem' / 'qa' /
'simeval_results.json')
cdd_res = read_results(testdata / 'nonmem' / 'qa' / 'cdd_results.json')
calculate_results(orig,
base,
simeval_results=simeval_res,
cdd_results=cdd_res)
def test_add_random_variables_and_statements(pheno_path):
model = Model(pheno_path)
rvs = model.random_variables
pset = model.parameters
eta = RandomVariable.normal('ETA_NEW', 'iiv', 0, S('omega'))
rvs.append(eta)
pset.append(Parameter('omega', 0.1))
eps = RandomVariable.normal('EPS_NEW', 'ruv', 0, S('sigma'))
rvs.append(eps)
pset.append(Parameter('sigma', 0.1))
model.random_variables = rvs
model.parameters = pset
sset = model.get_pred_pk_record().statements
statement_new = Assignment(S('X'), 1 + S(eps.name) + S(eta.name))
sset.append(statement_new)
model.get_pred_pk_record().statements = sset
model.update_source()
assert str(model.get_pred_pk_record()).endswith('X = 1 + ETA(3) + EPS(2)\n\n')
def test_special_models(testdata):
onePROB = testdata / 'nonmem' / 'modelfit_results' / 'onePROB'
withBayes = Model(onePROB / 'multEST' / 'noSIM' / 'withBayes.mod')
assert (pytest.approx(
withBayes.modelfit_results.standard_errors['THETA(1)'],
1e-13) == 2.51942e00)
assert (pytest.approx(
withBayes.modelfit_results[0].standard_errors['THETA(1)'],
1e-13) == 3.76048e-01)
assert withBayes.modelfit_results[0].minimization_successful is False
assert withBayes.modelfit_results[1].minimization_successful is False
assert withBayes.modelfit_results[0].covariance_step == {
'requested': True,
'completed': True,
'warnings': False,
}
assert withBayes.modelfit_results.covariance_step == {
'requested': True,
'completed': True,
'warnings': False,
}
maxeval0 = Model(onePROB / 'oneEST' / 'noSIM' / 'maxeval0.mod')
assert maxeval0.modelfit_results.minimization_successful is None
maxeval3 = Model(onePROB / 'oneEST' / 'noSIM' / 'maxeval3.mod')
assert maxeval3.modelfit_results.minimization_successful is False
assert maxeval3.modelfit_results.covariance_step == {
'requested': True,
'completed': True,
'warnings': True,
}
nearbound = Model(onePROB / 'oneEST' / 'noSIM' / 'near_bounds.mod')
correct = pd.Series(
[
False, True, False, False, False, False, False, False, True, True,
False
],
index=[
'THETA(1)',
'THETA(2)',
'THETA(3)',
'THETA(4)',
'OMEGA(1,1)',
'OMEGA(2,1)',
'OMEGA(2,2)',
'OMEGA(3,3)',
'OMEGA(4,4)',
'OMEGA(6,6)',
'SIGMA(1,1)',
],
)
pd.testing.assert_series_equal(nearbound.modelfit_results.near_bounds(),
correct)
def test_individual_parameter_statistics(testdata):
model = Model(testdata / 'nonmem' / 'secondary_parameters' / 'pheno.mod')
np.random.seed(103)
stats = individual_parameter_statistics(model, 'CL/V')
assert stats['mean'] == pytest.approx(0.00470525776968202)
assert stats['variance'] == pytest.approx(8.12398122254498e-6)
assert stats['stderr'] == pytest.approx(0.00344872, abs=1e-5)
model = Model(testdata / 'nonmem' / 'secondary_parameters' / 'run1.mod')
np.random.seed(5678)
def test_ofv(testdata):
base = Model(testdata / 'nonmem' / 'pheno.mod')
lin = Model(testdata / 'nonmem' / 'qa' / 'pheno_linbase.mod')
res = calculate_results(base, lin)
correct = """,ofv
base,730.894727
lin_evaluated,730.894727
lin_estimated,730.847272
"""
correct = pd.read_csv(StringIO(correct), index_col=[0])
pd.testing.assert_frame_equal(res.ofv, correct, atol=1e-6)
def update_model3b_for_psn(rundir, ncovs):
"""Function to update model3b from psn
NOTE: This function lets pharmpy tie in to the PsN workflow
and is a temporary solution
"""
model_path = Path(rundir) / 'm1'
model1b = Model(model_path / 'model_1b.mod')
model3 = Model(model_path / 'model_3.mod')
model3b = create_model3b(model1b, model3, int(ncovs))
model3b.write(model_path, force=True)
def test_transit_compartments_added_mdt(testdata):
model = Model(testdata / 'nonmem' / 'modeling' /
'pheno_advan5_nodepot.mod')
set_transit_compartments(model, 2)
transits = model.statements.ode_system.find_transit_compartments(
model.statements)
assert len(transits) == 2
model.update_source()
correct = ("""$PROBLEM PHENOBARB SIMPLE MODEL
$DATA ../pheno.dta IGNORE=@
$INPUT ID TIME AMT WGT APGR DV FA1 FA2
$SUBROUTINE ADVAN5 TRANS1
$MODEL COMPARTMENT=(TRANSIT1 DEFDOSE) COMPARTMENT=(TRANSIT2) COMPARTMENT=(CENTRAL) """
+ """COMPARTMENT=(PERIPHERAL)
$PK
MDT = THETA(6)
IF(AMT.GT.0) BTIME=TIME
TAD=TIME-BTIME
TVCL=THETA(1)*WGT
TVV=THETA(2)*WGT
IF(APGR.LT.5) TVV=TVV*(1+THETA(3))
CL=TVCL*EXP(ETA(1))
V=TVV*EXP(ETA(2))
K30 = CL/V
K34 = THETA(4)
K43 = THETA(5)
S3 = V
K12 = 2/MDT
K23 = 2/MDT
$ERROR
W=F
Y=F+W*EPS(1)
IPRED=F
IRES=DV-IPRED
IWRES=IRES/W
$THETA (0,0.00469307) ; CL
$THETA (0,1.00916) ; V
$THETA (-.99,.1)
$THETA (0,10)
$THETA (0,10)
$THETA (0,0.5) ; POP_MDT
$OMEGA DIAGONAL(2)
0.0309626 ; IVCL
0.031128 ; IVV
$SIGMA 1e-7
$ESTIMATION METHOD=1 INTERACTION
$COVARIANCE UNCONDITIONAL
$TABLE ID TIME DV AMT WGT APGR IPRED PRED RES TAD CWRES NPDE NOAPPEND
NOPRINT ONEHEADER FILE=sdtab1
""")
assert str(model) == correct
def test_initial_individual_estimates(datadir):
path = datadir / 'minimal.mod'
model = Model(path)
assert model.initial_individual_estimates is None
path = datadir / 'pheno_etas.mod'
model = Model(path)
inits = model.initial_individual_estimates
assert len(inits) == 59
assert len(inits.columns) == 2
assert inits['ETA(1)'][2] == -0.166321
