def test_updating_q(mcmc_initial_state):
rand = np.random.RandomState(2020)
q_sum = 0.0
for i in range(1000):
AlgorithmSiteInstance._update_q(mcmc_initial_state.state, rand)
q_sum += mcmc_initial_state.state.qq
q_average = q_sum / 1000.0
# TODO: Find a more robust way of testing this (since there's still a tiny chance it randomly falls outside this range?)
assert 0.375 <= q_average <= 0.380
def test_updating_dvect(mcmc_initial_state):
rand = np.random.RandomState(2020)
dposterior_sum = 0
dvect_sum = np.zeros(133)
for i in range(1000):
AlgorithmSiteInstance._update_dvect(mcmc_initial_state.state, rand)
dposterior_sum += mcmc_initial_state.state.dposterior
dvect_sum += mcmc_initial_state.state.dvect
dpost_average = dposterior_sum / 1000.0
dvect_average = dvect_sum / 1000.0
assert 0.31 <= dpost_average <= 0.32
np.testing.assert_approx_equal(dvect_average[0], dpost_average)
assert 0.202 <= dvect_average[1] <= 0.206
def _initialize_site_instances(self, genotypedata: pd.DataFrame,
additional: pd.DataFrame,
locirepeats: List[int]):
'''
Initializes the algorithm instances for each site in the processed
data file
:param locirepeats: TODO:
'''
self.site_names = pd.unique(genotypedata['Site'])
self.algorithm_instances = []
for site_name in self.site_names:
# NOTE: "RR" stands for "recrudescence and/or reinfection"; it marks
# datasets that deals specifically with day 0/day of failure info,
# as opposed to background data
site_genotypedata_RR = self._get_samples_from_site(
genotypedata, site_name)
site_additional_neutral = self._get_samples_from_site(
additional, site_name)
self._replace_sample_names(site_additional_neutral, 'Additional_')
site_instance = AlgorithmSiteInstance(site_genotypedata_RR,
site_additional_neutral,
locirepeats)
self.algorithm_instances.append((site_name, site_instance))
def test_likelihood_ratio_inner_loop_initial(mcmc_initial_state):
expected_inner_values = np.array([8.265306, 8.265306, 8.265306, 8.265306])
inner_values = AlgorithmSiteInstance._likelihood_inner_loop(
mcmc_initial_state.state, 0)
np.testing.assert_array_almost_equal(inner_values[0],
expected_inner_values,
decimal=5)
def test_likelihood_ratio(mcmc_initial_state):
expected_likelihood_ratios = np.array([56.024203, 1.438889, 0.0])
likelihood_ratios = AlgorithmSiteInstance._likelihood_ratios(
mcmc_initial_state.state, mcmc_initial_state.num_ids,
mcmc_initial_state.num_loci)
np.testing.assert_array_almost_equal(likelihood_ratios,
expected_likelihood_ratios,
decimal=5)
def test_likelihood_ratio_inner_loop_middle(mcmc_initial_state):
expected_inner_values = np.array(
[0.7777778, 0.7777778, 0.7777778, 0.7777778])
inner_values = AlgorithmSiteInstance._likelihood_inner_loop(
mcmc_initial_state.state, 2)
np.testing.assert_array_almost_equal(inner_values[1],
expected_inner_values,
decimal=5)
def test_likelihood_ratio_with_nans(mcmc_initial_state):
expected_likelihood_ratios = np.array([88.459268, 1.438889, 0.0])
mcmc_initial_state.state.alldistance[0, 1, 2:] = np.nan
mcmc_initial_state.state.allrecrf[0, 1, 2:] = np.nan
likelihood_ratios = AlgorithmSiteInstance._likelihood_ratios(
mcmc_initial_state.state, mcmc_initial_state.num_ids,
mcmc_initial_state.num_loci)
np.testing.assert_array_almost_equal(likelihood_ratios,
expected_likelihood_ratios,
decimal=5)
def test_updating_classifications(mcmc_initial_state):
expected_avg_classifications = np.array([1.0, 0.31, 0.0])
original_classifications = np.copy(mcmc_initial_state.state.classification)
classifications_sum = np.zeros(original_classifications.size)
likelihood_ratios = np.array([56.024203, 1.438889, 0.0])
rand = np.random.RandomState(2020)
for i in range(1000):
classifications = AlgorithmSiteInstance._update_classifications(
mcmc_initial_state.state, likelihood_ratios,
mcmc_initial_state.num_ids, rand)
classifications_sum += classifications
mcmc_initial_state.state.classification = original_classifications
avg_classes = classifications_sum / 1000.0
# TODO: Find a more robust way of testing this (since there's still a tiny chance it randomly falls outside this range?)
np.testing.assert_array_almost_equal(avg_classes,
expected_avg_classifications,
decimal=2)
def test_max_MOI():
maxMOI = AlgorithmSiteInstance._get_max_MOI(genotypedata_RR)
assert maxMOI == expected_maxMOI
example_file = os.path.join(os.path.dirname(__file__),
'../Angola2017_example.xlsx')
genotypedata, additional = RecrudescenceFileParser.parse_file(example_file)
genotypedata_RR = AlgorithmInstance._get_samples_from_site(
genotypedata, 'Benguela')
additional_neutral = AlgorithmInstance._replace_sample_names(
AlgorithmInstance._get_samples_from_site(additional, 'Benguela'),
'Additional_')
expected_maxMOI = 5
locirepeats = np.array([2, 2, 3, 3, 3, 3, 3])
expected_ids = pd.unique(
["BQ17-269", "BD17-040", "BD17-083", "BD17-085", "BD17-087", "BD17-090"])
expected_locinames = pd.unique(
["313", "383", "TA1", "POLYA", "PFPK2", "2490", "TA109"])
alleles_definitions_RR = AlgorithmSiteInstance._get_allele_definitions(
genotypedata_RR, additional_neutral, expected_locinames.size, locirepeats)
def test_max_MOI():
maxMOI = AlgorithmSiteInstance._get_max_MOI(genotypedata_RR)
assert maxMOI == expected_maxMOI
def test_getting_ids():
ids = recrudescence_utils.get_sample_ids(genotypedata_RR, 'Day 0')
np.testing.assert_array_equal(ids, expected_ids)
def test_getting_locinames():
locinames = pd.unique(genotypedata_RR.columns[1:].str.split("_").str[0])
def test_updating_frequencies(mcmc_initial_state):
rand = np.random.RandomState(2020)
AlgorithmSiteInstance._update_frequencies(mcmc_initial_state.state,
mcmc_initial_state.num_loci,
mcmc_initial_state.max_MOI, rand)