def test_insert_hints_overlapping() -> None:
"""Asserts that insert_hints works as expected for overlapping entities."""
text = (
"Mutation pattern in clinically asymptomatic coagulation factor VII deficiency. A total of"
" 122 subjects, referred after presurgery screening or checkup for prolonged prothrombin"
" time, were characterized for the presence of coagulation factor VII deficiency."
)
pubator_annotation = schemas.PubtatorAnnotation(
pmid="8844208",
text=text,
entities={
"2155":
schemas.PubtatorEntity(
mentions=["coagulation factor VII", "coagulation factor VII"],
offsets=[(44, 67), (222, 244)],
label="Gene",
),
"D005168":
schemas.PubtatorEntity(
mentions=["factor VII deficiency", "factor VII deficiency"],
offsets=[(56, 78), (234, 255)],
label="Disease",
),
},
)
expected = (
f"coagulation factor vii @GENE@ factor vii deficiency @DISEASE@ {HINT_SEP_SYMBOL} {text}"
)
pubator_annotation.insert_hints()
actual = pubator_annotation.text
assert actual == expected
def test_insert_hints_no_mutation() -> None:
"""Asserts that insert_hints does not mutate any attribute beside `text`."""
text = (
"Different lobular distributions of altered hepatocyte tight junctions in rat models of"
" intrahepatic and extrahepatic cholestasis.")
pubator_annotation = schemas.PubtatorAnnotation(
pmid="9862868",
text=text,
entities={
"D002780":
schemas.PubtatorEntity(mentions=["intrahepatic cholestasis"],
offsets=[(87, 128)],
label="Disease"),
"D001651":
schemas.PubtatorEntity(mentions=["extrahepatic cholestasis"],
offsets=[(104, 128)],
label="Disease"),
},
)
expected = copy.deepcopy(pubator_annotation)
pubator_annotation.insert_hints()
assert pubator_annotation.text != expected.text
assert pubator_annotation.pmid == expected.pmid
assert pubator_annotation.entities == expected.entities
assert pubator_annotation.relations == expected.relations
def test_insert_hints_compound() -> None:
"""Asserts that insert_hints works as expected for compound entities."""
text = (
"Different lobular distributions of altered hepatocyte tight junctions in rat models of"
" intrahepatic and extrahepatic cholestasis.")
pubator_annotation = schemas.PubtatorAnnotation(
pmid="9862868",
text=text,
entities={
"D002780":
schemas.PubtatorEntity(mentions=["intrahepatic cholestasis"],
offsets=[(87, 128)],
label="Disease"),
"D001651":
schemas.PubtatorEntity(mentions=["extrahepatic cholestasis"],
offsets=[(104, 128)],
label="Disease"),
},
)
expected = f"intrahepatic cholestasis @DISEASE@ extrahepatic cholestasis @DISEASE@ {HINT_SEP_SYMBOL} {text}"
pubator_annotation.insert_hints()
actual = pubator_annotation.text
assert actual == expected
def test_parse_pubtator_compound_ent() -> None:
# A truncated example taken from the CDR dataset
pmid = "17854040"
title_text = (
"Mutations associated with lamivudine-resistance in therapy-na ve hepatitis B virus (HBV)"
" infected patients with and without HIV co-infection: implications for antiretroviral"
" therapy in HBV and HIV co-infected South African patients. infected patients with and"
" without HIV co-infection: implications for antiretroviral therapy in HBV and HIV"
" co-infected South African patients.")
abstract_text = (
"This was an exploratory study to investigate lamivudine-resistant hepatitis B virus (HBV)"
" strains in selected lamivudine-na ve HBV carriers with and without human"
" immunodeficiency virus (HIV) co-infection in South African patients. Thirty-five"
" lamivudine-naive HBV infected patients with or without HIV co-infection were studied: 15"
" chronic HBV mono-infected patients and 20 HBV-HIV co-infected patients."
)
pubtator_content = f"""
{pmid}|t|{title_text}
{pmid}|a|{abstract_text}
{pmid}\t26\t36\tlamivudine\tChemical\tD019259
{pmid}\t59\t61\tna\tChemical\tD012964
{pmid}\t66\t98\thepatitis B virus (HBV) infected\tDisease\tD006509
{pmid}\t125\t141\tHIV co-infection\tDisease\tD015658
{pmid}\t186\t209\tHBV and HIV co-infected\tDisease\tD006509|D015658 HBV infected|HIV infected
"""
expected = schemas.PubtatorAnnotation(
pmid=pmid,
text=f"{title_text} {abstract_text}",
entities={
"D019259":
schemas.PubtatorEntity(
mentions=["lamivudine"],
offsets=[(26, 36)],
label="Chemical",
),
"D012964":
schemas.PubtatorEntity(mentions=["na"],
offsets=[(59, 61)],
label="Chemical"),
"D006509":
schemas.PubtatorEntity(
mentions=["hepatitis B virus (HBV) infected", "HBV infected"],
offsets=[(66, 98), (186, 209)],
label="Disease",
),
"D015658":
schemas.PubtatorEntity(
mentions=["HIV co-infection", "HIV infected"],
offsets=[(125, 141), (194, 209)],
label="Disease",
),
},
)
actual = util.parse_pubtator(pubtator_content,
text_segment=util.TextSegment.both)
assert actual[0].text == expected.text
assert actual[0].entities == expected.entities
assert actual[0].relations == expected.relations
def test_pubtator_entity_to_string() -> None:
ent = schemas.PubtatorEntity(
# Contains:
# - multi-word mentions
# - overlapping mentions
# - multiple duplicate mentions
# - at least two unique mentions (case-insensitive)
# - mentions that are not already ordered by first appearance
mentions=[
"factor vii deficiency",
"factor vii deficiency",
"Factor VII Deficiency",
"factor vii deficient",
],
offsets=[(200, 221), (100, 121), (20, 41), (0, 21)],
label="Disease",
)
# Test with sorting (which is and should be the default)
actual = ent.to_string()
expected = f"factor vii deficient {COREF_SEP_SYMBOL} factor vii deficiency @DISEASE@"
assert actual == expected
# Test without sorting
# Note: because the mentions are randomly sorted when sort=False, we check a couple other
# attributes, like length of the string.
actual = ent.to_string(sort=False)
assert len(actual) == len(expected)
assert "factor vii deficient" in actual
assert "factor vii deficiency" in actual
assert "@DISEASE@" in actual
assert COREF_SEP_SYMBOL in actual
def test_insert_hints() -> None:
"""Asserts that insert_hints works as expected for a list of edge cases."""
# A truncated example taken from the GDA dataset. It contains a few edge cases:
# - coreferent mention
# - entites that differ in case
# - paranthesized entity
# - multiple identical mentions of an entity
text = (
"Apolipoprotein E epsilon4 allele, elevated midlife total cholesterol level, and high"
" midlife systolic blood pressure are independent risk factors for late-life Alzheimer disease."
" BACKGROUND: Presence of the apolipoprotein E (apoE) epsilon4 allele, which is involved in"
" cholesterol metabolism, is the most important genetic risk factor for Alzheimer disease."
" Elevated midlife values for total cholesterol level and blood pressure have been"
" implicated recently as risk factors for Alzheimer disease.")
pubator_annotation = schemas.PubtatorAnnotation(
pmid="12160362",
text=text,
entities={
"348":
schemas.PubtatorEntity(
mentions=["Apolipoprotein E", "apolipoprotein E", "apoE"],
offsets=[(0, 17), (207, 223), (225, 229)],
label="Gene",
),
"D000544":
schemas.PubtatorEntity(
mentions=[
"Alzheimer disease", "Alzheimer disease",
"Alzheimer disease"
],
offsets=[(160, 177), (339, 356), (479, 496)],
label="Disease",
),
},
)
expected = f"apolipoprotein e {COREF_SEP_SYMBOL} apoe @GENE@ alzheimer disease @DISEASE@ {HINT_SEP_SYMBOL} {text}"
pubator_annotation.insert_hints()
actual = pubator_annotation.text
assert actual == expected
def test_filter_hypernyms(self):
annotation = schemas.PubtatorAnnotation(
text=(
"Carbamazepine-induced cardiac dysfunction. A patient with sinus bradycardia and"
" atrioventricular block, induced by carbamazepine, prompted an extensive"
" literature review of all previously reported cases."
),
pmid="",
entities={
"D002220": schemas.PubtatorEntity(
mentions=["Carbamazepine", "carbamazepine"],
offsets=[(0, 13), (115, 128)],
label="Chemical",
),
"D006331": schemas.PubtatorEntity(
mentions=["cardiac dysfunction"],
offsets=[(22, 41)],
label="Disease",
),
"D001919": schemas.PubtatorEntity(
mentions=["bradycardia"],
offsets=[(64, 75)],
label="Disease",
),
"D054537": schemas.PubtatorEntity(
mentions=["atrioventricular block"],
offsets=[(80, 102)],
label="Disease",
),
},
relations=[("D002220", "D001919", "CID"), ("D002220", "D054537", "CID")],
)
cdr._filter_hypernyms([annotation])
actual = annotation.filtered_relations
# D006331 is a hypernym of D001919 and/or D054537 and so it should be filtered.
expected = [("D002220", "D006331", "CID")]
assert actual == expected
def test_pubtator_entity_get_offsets() -> None:
ent = schemas.PubtatorEntity(
# We don't need actual mentions or a label to test this method.
mentions=[
"",
"",
"",
"",
],
offsets=[(200, 221), (100, 121), (20, 41), (0, 21)],
label="",
)
expected = (0, 21)
actual = ent.get_offsets()
assert actual == expected
def test_pubtator_annotation_to_string() -> None:
# Contains:
# - at least one entity with multiple mentions, including a unique mention
# - at least two relations with different head entities
# - at least one n-ary relation
# - relations that are not already ordered by first appearance
ann = schemas.PubtatorAnnotation(
# We don't need text or a PMID to test this method.
pmid="",
text="",
entities={
"D008094":
schemas.PubtatorEntity(
mentions=["lithium", "lithium", "Li", "Li"],
offsets=[(54, 61), (111, 118), (941, 943), (1333, 1335)],
label="Chemical",
),
"D006973":
schemas.PubtatorEntity(
mentions=["hypertension", "hypertension"],
offsets=[(1000, 1012), (1500, 1512)],
label="Disease",
),
"D011507":
schemas.PubtatorEntity(
mentions=["proteinuria", "proteinuria"],
offsets=[(975, 986), (1466, 1477)],
label="Disease",
),
"D007676":
schemas.PubtatorEntity(
mentions=["chronic renal failure", "chronic renal failure"],
offsets=[(70, 91), (1531, 1552)],
label="Disease",
),
},
relations=[
("D008094", "D006973", "CID"),
("D008094", "D011507", "CID"),
("D008094", "D007676", "CID"),
# This is an artificial n-ary relation.
("D008094", "D006973", "D011507", "CID"),
],
)
# Test with sorting (which is and should be the default)
actual = ann.to_string()
expected = (
f"lithium {COREF_SEP_SYMBOL} li @CHEMICAL@ chronic renal failure @DISEASE@ @CID@"
f" lithium {COREF_SEP_SYMBOL} li @CHEMICAL@ proteinuria @DISEASE@ @CID@"
f" lithium {COREF_SEP_SYMBOL} li @CHEMICAL@ hypertension @DISEASE@ @CID@"
f" lithium {COREF_SEP_SYMBOL} li @CHEMICAL@ hypertension @DISEASE@ proteinuria @DISEASE@ @CID@"
)
assert actual == expected
# Test without sorting
# Note: because the mentions are randomly sorted when sort=False, we check a couple other
# attributes, like length of the string.
actual = ann.to_string(sort=False)
assert len(actual) == len(expected)
assert "lithium" in actual
assert "li" in actual
assert "chronic renal failure" in actual
assert "proteinuria" in actual
assert "hypertension" in actual
assert "@CHEMICAL@" in actual
assert "@DISEASE@" in actual
assert "@CID@" in actual
assert COREF_SEP_SYMBOL in actual
def test_parse_pubtator() -> None:
# A truncated example taken from the CDR dataset
pmid = "18020536"
title_text = (
"Associations between use of benzodiazepines or related drugs and health, physical"
" abilities and cognitive function: a non-randomised clinical study in the elderly."
)
abstract_text = (
"OBJECTIVE: To describe associations between the use of benzodiazepines or related drugs"
" (BZDs/RDs) and health, functional abilities and cognitive function in the elderly."
" METHODS: A non-randomised clinical study of patients aged > or =65 years admitted to"
" acute hospital wards during 1 month. 164 patients (mean age +/- standard deviation [SD]"
" 81.6 +/- 6.8 years) were admitted. Of these, nearly half (n = 78) had used BZDs/RDs"
" before admission, and the remainder (n = 86) were non-users. Cognitive ability was"
" assessed by the Mini-Mental State Examination (MMSE). Patients scoring > or =20 MMSE"
" sum points were interviewed (n = 79) and questioned regarding symptoms and functional"
" abilities during the week prior to admission.")
# Include a dummy annotation with ID == -1. These should be ignored.
pubtator_content = f"""
{pmid}|t|{title_text}
{pmid}|a|{abstract_text}
{pmid}\t28\t43\tbenzodiazepines\tChemical\tD001569
{pmid}\t219\t234\tbenzodiazepines\tChemical\tD001569
{pmid}\t253\t257\tBZDs\tChemical\tD001569
{pmid}\t583\t587\tBZDs\tChemical\tD001569
{pmid}\t1817\t1826\ttiredness\tDisease\tD005221
{pmid}\t0\t0\tArbitrary\tArbitrary\t-1
{pmid}\tCID\tD001569\tD005221
"""
title_entities = {
"D001569":
schemas.PubtatorEntity(
mentions=["benzodiazepines"],
offsets=[(28, 43)],
label="Chemical",
),
}
abstract_entities = {
"D001569":
schemas.PubtatorEntity(
mentions=["benzodiazepines", "BZDs", "BZDs"],
offsets=[(219, 234), (253, 257), (583, 587)],
label="Chemical",
),
"D005221":
schemas.PubtatorEntity(mentions=["tiredness"],
offsets=[(1817, 1826)],
label="Disease"),
}
both_entities = {
"D001569":
schemas.PubtatorEntity(
mentions=title_entities["D001569"].mentions +
abstract_entities["D001569"].mentions,
offsets=title_entities["D001569"].offsets +
abstract_entities["D001569"].offsets,
label="Chemical",
),
"D005221":
schemas.PubtatorEntity(mentions=["tiredness"],
offsets=[(1817, 1826)],
label="Disease"),
}
# Title only
expected = schemas.PubtatorAnnotation(pmid=pmid,
text=title_text,
entities=title_entities,
relations=[])
actual = util.parse_pubtator(pubtator_content,
text_segment=util.TextSegment.title)
# Breaking up the asserts leads to much clearer outputs when the test fails
assert actual[0].text == expected.text
assert actual[0].entities == expected.entities
assert actual[0].relations == expected.relations
# Abstract only
expected = schemas.PubtatorAnnotation(
pmid=pmid,
text=abstract_text,
entities=abstract_entities,
relations=[("D001569", "D005221", "CID")],
)
actual = util.parse_pubtator(pubtator_content,
text_segment=util.TextSegment.abstract)
assert actual[0].text == expected.text
assert actual[0].entities == expected.entities
assert actual[0].relations == expected.relations
# Both
expected = schemas.PubtatorAnnotation(
pmid=pmid,
text=f"{title_text} {abstract_text}",
entities=both_entities,
relations=[("D001569", "D005221", "CID")],
)
actual = util.parse_pubtator(pubtator_content,
text_segment=util.TextSegment.both)
assert actual[0].text == expected.text
assert actual[0].entities == expected.entities
assert actual[0].relations == expected.relations
def test_query_pubtator() -> None:
pmid = "19285439"
title_text = (
"The ubiquitin ligase RNF5 regulates antiviral responses by mediating degradation"
" of the adaptor protein MITA.")
abstract_text = (
"Viral infection activates transcription factors NF-kappaB and IRF3, which collaborate to"
" induce type I interferons (IFNs) and elicit innate antiviral response. MITA (also known"
" as STING) has recently been identified as an adaptor that links virus-sensing receptors"
" to IRF3 activation. Here, we showed that the E3 ubiquitin ligase RNF5 interacted with"
" MITA in a viral-infection-dependent manner. Overexpression of RNF5 inhibited"
" virus-triggered IRF3 activation, IFNB1 expression, and cellular antiviral response,"
" whereas knockdown of RNF5 had opposite effects. RNF5 targeted MITA at Lys150 for"
" ubiquitination and degradation after viral infection. Both MITA and RNF5 were located at"
" the mitochondria and endoplasmic reticulum (ER) and viral infection caused their"
" redistribution to the ER and mitochondria, respectively. We further found that"
" virus-induced ubiquitination and degradation of MITA by RNF5 occurred at the"
" mitochondria. These findings suggest that RNF5 negatively regulates virus-triggered"
" signaling by targeting MITA for ubiquitination and degradation at the mitochondria."
)
title_entities = {
"6048":
schemas.PubtatorEntity(
mentions=["RNF5"],
offsets=[(21, 25)],
label="Gene",
),
"340061":
schemas.PubtatorEntity(mentions=["MITA"],
offsets=[(104, 108)],
label="Gene"),
}
abstract_entities = {
"4790":
schemas.PubtatorEntity(mentions=["NF-kappaB"],
offsets=[(158, 167)],
label="Gene"),
"3661":
schemas.PubtatorEntity(
mentions=["IRF3", "IRF3", "IRF3"],
offsets=[(172, 176), (378, 382), (554, 558)],
label="Gene",
),
"340061":
schemas.PubtatorEntity(
mentions=["MITA", "STING", "MITA", "MITA", "MITA", "MITA", "MITA"],
offsets=[
(270, 274),
(290, 295),
(461, 465),
(684, 688),
(762, 766),
(1000, 1004),
(1136, 1140),
],
label="Gene",
),
"6048":
schemas.PubtatorEntity(
mentions=["RNF5", "RNF5", "RNF5", "RNF5", "RNF5", "RNF5", "RNF5"],
offsets=[
(440, 444),
(523, 527),
(643, 647),
(670, 674),
(771, 775),
(1008, 1012),
(1071, 1075),
],
label="Gene",
),
"3456":
schemas.PubtatorEntity(mentions=["IFNB1"],
offsets=[(571, 576)],
label="Gene"),
}
both_entities = {
"6048":
schemas.PubtatorEntity(
mentions=title_entities["6048"].mentions +
abstract_entities["6048"].mentions,
offsets=title_entities["6048"].offsets +
abstract_entities["6048"].offsets,
label="Gene",
),
"340061":
schemas.PubtatorEntity(
mentions=title_entities["340061"].mentions +
abstract_entities["340061"].mentions,
offsets=title_entities["340061"].offsets +
abstract_entities["340061"].offsets,
label="Gene",
),
"4790":
schemas.PubtatorEntity(mentions=["NF-kappaB"],
offsets=[(158, 167)],
label="Gene"),
"3661":
schemas.PubtatorEntity(
mentions=abstract_entities["3661"].mentions,
offsets=abstract_entities["3661"].offsets,
label="Gene",
),
"3456":
schemas.PubtatorEntity(mentions=["IFNB1"],
offsets=[(571, 576)],
label="Gene"),
}
# Title only
expected = schemas.PubtatorAnnotation(pmid=pmid,
text=title_text,
entities=title_entities,
relations=[])
actual = util.query_pubtator(pmids=[pmid],
concepts=["gene"],
text_segment=util.TextSegment.title)
# Breaking up the asserts leads to much clearer outputs when the test fails
assert len(actual) == 1
assert actual[expected.pmid].text == expected.text
assert actual[expected.pmid].entities == expected.entities
assert actual[expected.pmid].relations == expected.relations
# Abstract only
expected = schemas.PubtatorAnnotation(pmid=pmid,
text=abstract_text,
entities=abstract_entities,
relations=[])
actual = util.query_pubtator(pmids=[pmid],
concepts=["gene"],
text_segment=util.TextSegment.abstract)
assert len(actual) == 1
assert actual[expected.pmid].text == expected.text
assert actual[expected.pmid].entities == expected.entities
assert actual[expected.pmid].relations == expected.relations
# Both
expected = schemas.PubtatorAnnotation(pmid=pmid,
text=f"{title_text} {abstract_text}",
entities=both_entities,
relations=[])
actual = util.query_pubtator(pmids=[pmid],
concepts=["gene"],
text_segment=util.TextSegment.both)
assert len(actual) == 1
assert actual[expected.pmid].text == expected.text
assert actual[expected.pmid].entities == expected.entities
assert actual[expected.pmid].relations == expected.relations