def calculate_area(nested_genotype: pandas.Series,
unnested_genotype: pandas.Series) -> float:
"""
Calculates a score based on the probability the unnested genotype is a subset of the nested_genotype.
This take into account both the area of the nested genotype and the area of all other genotypes not in the nested genotype.
Parameters
----------
nested_genotype, unnested_genotype: pandas.Series
"""
# If the nested genotype is not fixed, group the remaining frequencies into an `other` category.
# noinspection PyTypeChecker
other_genotypes: pandas.Series = 1 - nested_genotype
area_nested = areascore.area_of_series(nested_genotype)
area_unnested = areascore.area_of_series(unnested_genotype)
area_other = areascore.area_of_series(other_genotypes)
common_area_nested = areascore.calculate_common_area(
nested_genotype, unnested_genotype)
common_area_other = areascore.calculate_common_area(
other_genotypes, unnested_genotype)
is_subset_nested = areascore.is_subset_legacy(area_nested,
area_unnested,
common_area_nested)
is_subset_other = areascore.is_subset_legacy(area_other, area_unnested,
common_area_other)
if is_subset_nested and is_subset_other:
score = int(common_area_nested > 2 * common_area_other)
elif is_subset_nested:
score = 2
else:
score = -2
# logger.debug(f"{unnested_genotype.name}\t{nested_genotype.name}\t{score}")
return score
def test_area_of_series_real():
filename = list(real_tables.values())[0]
table_genotype = pandas.read_excel(
filename, sheet_name="genotype").set_index('Genotype')
for index, row in table_genotype.iterrows():
logger.debug(index)
expected_area = areascore.calculate_area(row)
actual_area = areascore.area_of_series(row)
assert pytest.approx(actual_area, abs=.001) == expected_area
def test_calculate_overlapping_area(full_overlap, no_overlap, partial_overlap):
left, right = full_overlap
right_area = areascore.area_of_series(right)
overlap_area = areascore.calculate_common_area(left, right)
assert pytest.approx(overlap_area, abs=0.001) == right_area
left, right = no_overlap
overlap_area = areascore.calculate_common_area(left, right)
assert pytest.approx(overlap_area, abs=0.001) == 0
left, right = partial_overlap
overlap_area = areascore.calculate_common_area(left, right)
assert pytest.approx(overlap_area, abs=0.001) == right_area
left = pandas.Series([
0.01, 0.279, 0.341, 0.568, 0.708, 0.913, 0.756, 0.455, 0.399, 0.13,
0.041
])
right = pandas.Series(
[0, 0, 0, 0, 0, 0.247, 0.388, 0.215, 0.399, 0.13, 0.028])
overlap_area = areascore.calculate_common_area(left, right)
assert pytest.approx(overlap_area) == areascore.area_of_series(right)
def test_area_of_series(series, expected):
s = pandas.Series(series)
result = areascore.area_of_series(s)
assert pytest.approx(result, expected)
def jaccard_distance(left: pandas.Series, right: pandas.Series) -> float:
area_left = area_of_series(left)
area_right = area_of_series(right)
area_shared = calculate_common_area(left, right)
j = area_shared / (area_left + area_right - area_shared)
return 1 - j
def test_shoelace(trajectory_table, key, expected): # Keep this for now to make sure the area is being calculated correctly. series = trajectory_table.loc[key] result = areascore.area_of_series(series) assert pytest.approx(result, abs = 0.01) == expected
def calculate_score_area(self, nested_genotype: pandas.Series,
unnested_genotype: pandas.Series) -> float:
"""
Calculates a score based on the probability the unnested genotype is a subset of the nested_genotype.
This take into account both the area of the nested genotype and the area of all other genotypes not in the nested genotype.
Parameters
----------
nested_genotype, unnested_genotype: pandas.Series
"""
# If the nested genotype is not fixed, group the remaining frequencies into an `other` category.
difference_series = nested_genotype - unnested_genotype
if difference_series.mean() > 0:
# noinspection PyTypeChecker
other_genotypes: pandas.Series = self.flimit - nested_genotype
else:
# noinspection PyTypeChecker
other_genotypes: pandas.Series = self.flimit - unnested_genotype # In case we're testing if a small genotype contains a large genotype
other_genotypes = other_genotypes.mask(
lambda s: s < 0, 0.0001) # Since the flimit is not exactly 1.
unnested_polygon = polygon.as_polygon(unnested_genotype)
nested_polygon = polygon.as_polygon(nested_genotype)
other_polygon = polygon.as_polygon(other_genotypes)
is_subset_nested = areascore.is_subset_polygon(nested_polygon,
unnested_polygon)
is_subset_other = areascore.is_subset_polygon(other_polygon,
unnested_polygon)
is_subset_nested_reversed = areascore.is_subset_polygon(
unnested_polygon, nested_polygon) # Check the reverse case
nested_area = areascore.area_of_series(nested_genotype)
unnested_area = areascore.area_of_series(unnested_genotype)
common_area_nested = areascore.X_and_Y_polygon(unnested_polygon,
nested_polygon)
xor_area_unnested = areascore.difference_polygon(
unnested_polygon, nested_polygon
) # This does not distinguish between xor left vs xor right
if self.debug:
logger.debug(
f"calculate_score_jaccard()->({is_subset_nested}, {is_subset_other}, {is_subset_nested_reversed}), ({common_area_nested:.2f}, {xor_area_unnested:.2f})"
)
if is_subset_nested and is_subset_other:
# Evidence for both scenarios
# Test if the nested genotype is sufficiently large to assume the unnested genotype is a subset.
# Test only the area where the unnested genotype was detected.
other_polygon = polygon.as_polygon(other_genotypes)
common_area_other = areascore.X_and_Y_polygon(
unnested_polygon, other_polygon)
score = int(common_area_nested > 2 * common_area_other)
elif is_subset_nested:
score = 1
elif is_subset_nested_reversed and not is_subset_other:
score = -1
else:
score = 0
if score == 0 and xor_area_unnested > common_area_nested * 2:
score = -1
elif unnested_area > 2 * nested_area:
score = -1
score = score * self.weight_jaccard
return score