def test_search(mock_fetch_formulas, spark_context, ds_config):
formulas = ['H2O', 'C5H3O']
adducts = ds_config['isotope_generation']['adducts']
mock_fetch_formulas.side_effect = lambda moldb_id: formulas
with TemporaryDirectory() as tmpdir:
ds_data_path = Path(tmpdir)
msm_search = MSMSearch(
spark_context,
make_imzml_reader_mock(),
[MolecularDB(0, 'tests_db', 'version', targeted=True)],
ds_config,
ds_data_path,
NullProfiler(),
)
msm_search._fetch_formula_centroids = make_fetch_formula_centroids_mock()
msm_search.process_segments = lambda centr_segm_n, func: spark_context.parallelize(
map(func, range(centr_segm_n))
)
moldb_ion_metrics_df, moldb_ion_images_rdd, fdr_bundle = next(msm_search.search())
assert len(moldb_ion_metrics_df) == len(formulas) * len(adducts)
assert moldb_ion_images_rdd.count() == len(formulas) * len(adducts)
sanity_check_fdr_diagnostics(fdr_bundle)
def test_imzml_reader_tic_image_from_spectra():
imzml_reader = make_imzml_reader_mock(coordinates=MOCK_COORDINATES, spectra=MOCK_SPECTRA)
# Ensure all spectra have been read before getting the TIC image
for _ in imzml_reader.iter_spectra(np.arange(4)):
pass
assert np.array_equal(imzml_reader.tic_image(), EXPECTED_TIC, equal_nan=True)
def test_segment_ds(dump_mock):
imzml_reader = make_imzml_reader_mock(list(product(
[0], range(10))), (np.linspace(0, 90, num=10), np.ones(10)))
ds_segments = np.array([[0, 50], [50, 90.0]])
chunk_sp_n = 1000
segment_ds(imzml_reader, chunk_sp_n, ds_segments, Path('/tmp/abc'))
for segm_i, ((sp_chunk_df, f), _) in enumerate(dump_mock.call_args_list):
min_mz, max_mz = ds_segments[segm_i]
assert sp_chunk_df.shape == (50, 3)
assert np.all(min_mz <= sp_chunk_df.mz)
assert np.all(sp_chunk_df.mz <= max_mz)
def _test_compute_metrics():
ds_config = DSConfig(
analysis_version=1,
image_generation=DSConfigImageGeneration(ppm=3.0,
n_levels=30,
min_px=1,
compute_unused_metrics=False),
# Unused fields
database_ids=None,
isotope_generation=None,
fdr=None,
)
imzml_reader = make_imzml_reader_mock(list(product(range(2), range(3))))
compute_metrics = make_compute_image_metrics(imzml_reader, ds_config)
return compute_metrics
def test_fetch_chunk_spectra_data():
mz_n = 10
imzml_reader = make_imzml_reader_mock(
[(1, 1, 1), (2, 1, 1)], (np.linspace(0, 90, num=mz_n), np.ones(mz_n)))
sp_chunk_df = fetch_chunk_spectra_data(sp_ids=[0, 1],
imzml_reader=imzml_reader)
exp_mzs, exp_ints = [
np.sort([mz for mz in np.linspace(0, 90, num=mz_n) for _ in range(2)]),
np.ones(2 * mz_n),
]
assert sp_chunk_df.mz.dtype == 'f'
assert_array_almost_equal(sp_chunk_df.mz, exp_mzs)
assert_array_almost_equal(sp_chunk_df.int, exp_ints)
def test_imzml_reader_tic_image_from_metadata():
imzml_reader = make_imzml_reader_mock(
coordinates=MOCK_COORDINATES,
spectra=MOCK_SPECTRA,
spectrum_metadata_fields={TIC_ACCESSION: [0, 1, 2, 3]},
)
# metadata_tic intentionally matches the metadata but not the spectra
metadata_tic = np.array(
[
[0, np.nan, 1],
[np.nan, np.nan, np.nan],
[2, np.nan, 3],
]
)
assert np.array_equal(imzml_reader.tic_image(), metadata_tic, equal_nan=True)
def test_define_ds_segments():
imzml_reader = make_imzml_reader_mock(mz_precision='d')
mz_max = 100
sample_mzs = np.linspace(0, mz_max, 100)
ds_segm_size_mb = 800 / (
2**20) # 1600 b total data size / 2 segments, converted to MB
ds_segments = define_ds_segments(sample_mzs,
sample_ratio=1,
imzml_reader=imzml_reader,
ds_segm_size_mb=ds_segm_size_mb)
exp_ds_segm_n = 8
exp_bounds = [i * mz_max / exp_ds_segm_n for i in range(exp_ds_segm_n + 1)]
exp_ds_segments = np.array(list(zip(exp_bounds[:-1], exp_bounds[1:])))
assert ds_segments.shape == exp_ds_segments.shape
assert np.allclose(ds_segments, exp_ds_segments)
def test_ambiguous_modifiers(
fetch_formulas_mock, formula_image_metrics_mock, spark_context, ds_config
):
with TemporaryDirectory() as tmpdir:
ds_data_path = Path(tmpdir)
print(ds_data_path)
ds_config = {
**ds_config,
"isotope_generation": {
**ds_config["isotope_generation"],
# This set of modifiers are deliberately chosen so that ('','-H2O','+H') and ('-H2O+H','','') produce the same
# modifier string, to test that no code accidentally relies on "modifier" or "ion" strings being unambiguous
"chem_mods": ["-H2O+H"],
"neutral_losses": ["-H2O"],
"adducts": ["+H", "[M]+"],
},
}
formulas = [
'H3O',
'H4O',
'H5O2',
'H6O2',
] # Formulae selected to create isomers with the above modifiers
fetch_formulas_mock.return_value = formulas
msm_search = MSMSearch(
spark_context,
make_imzml_reader_mock(),
[MolecularDB(0, 'test_db', 'version', targeted=True)],
ds_config,
ds_data_path,
NullProfiler(),
)
msm_search._fetch_formula_centroids = make_fetch_formula_centroids_mock()
msm_search.process_segments = lambda centr_segm_n, func: spark_context.parallelize(
map(func, range(centr_segm_n))
)
formula_image_metrics_mock.side_effect = make_formula_image_metrics_mock_side_effect()
moldb_ion_metrics_df, _, fdr_bundle = next(msm_search.search())
assert (
moldb_ion_metrics_df[['formula', 'chem_mod', 'neutral_loss', 'adduct']]
.duplicated()
.sum()
== 0
)
# There are 4 combinations of modifiers to get H2: (H3O,-H2O+H,,), (H3O,,-H2O,+H), (H4O,,-H2O,), (H5O2,-H2O+H,-H2O,)
assert len(moldb_ion_metrics_df[moldb_ion_metrics_df.ion_formula == 'H2']) == 4
# Only 1 combination of modifiers can create H7O2: (H6O2,,,+H)
assert len(moldb_ion_metrics_df[moldb_ion_metrics_df.ion_formula == 'H7O2']) == 1
# H5O2 and H6O2 can have all combinations: 2 neutral loss options, 2 chem mods, 2 adducts = 8 possible combinations
assert len(moldb_ion_metrics_df[moldb_ion_metrics_df.formula == 'H5O2']) == 8
assert len(moldb_ion_metrics_df[moldb_ion_metrics_df.formula == 'H6O2']) == 8
# H3O and H4O cannot simultaneously have -H2O and -H2O+H
assert len(moldb_ion_metrics_df[moldb_ion_metrics_df.formula == 'H3O']) == 6
assert len(moldb_ion_metrics_df[moldb_ion_metrics_df.formula == 'H4O']) == 6
assert moldb_ion_metrics_df.formula.isin(formulas).all()
sanity_check_fdr_diagnostics(fdr_bundle)