Merge pull request #411 from apriltuesday/automated-mappings

Improve automated Zooma mappings

cmat/trait_mapping/main.py

@@ -4,10 +4,11 @@
 from collections import Counter
 
 from cmat.clinvar_xml_io import ClinVarTrait
+from cmat.trait_mapping.ols import get_uri_from_exact_match
 from cmat.trait_mapping.output import output_trait
 from cmat.trait_mapping.oxo import get_oxo_results
 from cmat.trait_mapping.oxo import uris_to_oxo_format
-from cmat.trait_mapping.trait import Trait
+from cmat.trait_mapping.trait import Trait, OntologyEntry
 from cmat.trait_mapping.trait_names_parsing import parse_trait_names
 from cmat.trait_mapping.zooma import get_zooma_results
 
@@ -32,8 +33,9 @@ def get_uris_for_oxo(zooma_result_list: list) -> set:
 
 def process_trait(trait: Trait, filters: dict, zooma_host: str, oxo_target_list: list, oxo_distance: int, target_ontology: str = 'EFO') -> Trait:
     """
-    Process a single trait. Find any mappings in Zooma. If there are no high confidence Zooma
-    mappings that are in EFO then query OxO with any high confidence mappings not in EFO.
+    Process a single trait. First look for an exact string match in the target ontology and return immediately if found.
+    Otherwise find any mappings in Zooma. If there are no high confidence Zooma mappings that are in EFO then query OxO
+    with any high confidence mappings not in EFO.
 
     :param trait: The trait to be processed.
     :param filters: A dictionary of filters to use for querying Zooma.
@@ -47,7 +49,12 @@ def process_trait(trait: Trait, filters: dict, zooma_host: str, oxo_target_list:
     """
     logger.debug('Processing trait {}'.format(trait.name))
 
-    trait.zooma_result_list = get_zooma_results(trait.name, filters, zooma_host, target_ontology)
+    string_match_uri = get_uri_from_exact_match(trait.name.lower(), target_ontology)
+    if string_match_uri:
+        trait.finished_mapping_set.add(OntologyEntry(string_match_uri, trait.name.lower()))
+        return trait
+
+    trait.zooma_result_list = get_zooma_results(trait.name.lower(), filters, zooma_host, target_ontology)
     trait.process_zooma_results()
     if (trait.is_finished
             or len(trait.zooma_result_list) == 0

cmat/trait_mapping/ols.py

@@ -1,3 +1,4 @@
+import os
 from functools import lru_cache
 import logging
 import requests
@@ -121,3 +122,31 @@ def get_replacement_term(uri: str, ontology: str = 'EFO') -> str:
     if response_json["term_replaced_by"] is not None:
         return response_json["term_replaced_by"]
     return ""
+
+
+@lru_cache(maxsize=16384)
+@retry(exceptions=(ConnectionError, requests.RequestException), tries=4, delay=2, backoff=1.2, jitter=(1, 3))
+def get_uri_from_exact_match(text, ontology='EFO'):
+    """
+    Finds URI from target ontology for a given text based on exact string match.
+
+    :param text: String to search for
+    :param ontology: ID of target ontology to query (default EFO)
+    :return: URI of matching term or None if not found
+    """
+    search_url = os.path.join(OLS_SERVER, f'api/search?ontology={ontology}&q={text}&queryFields=label&exact=true')
+    response = requests.get(search_url)
+    response.raise_for_status()
+    data = response.json()
+    if 'response' in data:
+        results = data['response']['docs']
+        candidates = set()
+        for result in results:
+            # Check that we've found the term exactly (strict case-insensitive string match)
+            if result['label'].lower() == text.lower():
+                candidates.add(result['iri'])
+        # Only return a result if we can find it unambiguously
+        if len(candidates) == 1:
+            return candidates.pop()
+    logger.warning(f'Could not find an IRI for {text}')
+    return None

cmat/trait_mapping/trait.py

@@ -58,12 +58,11 @@ def process_zooma_results(self):
         Check whether any Zooma mappings can be output as a finished ontology mapping.
         Put any finished mappings in finished_mapping_set
         """
-        for mapping in self.zooma_result_list:
-            if mapping.confidence.lower() != "high":
-                continue
-
-            for mapping in mapping.mapping_list:
-                if mapping.in_ontology and mapping.is_current:
+        for zooma_result in self.zooma_result_list:
+            for mapping in zooma_result.mapping_list:
+                # Accept current mappings in the target ontology with either high-confidence or exact string matches
+                if mapping.in_ontology and mapping.is_current and (zooma_result.confidence.lower() == "high"
+                                                                   or zooma_result.zooma_label.lower() == self.name.lower()):
                     ontology_entry = OntologyEntry(mapping.uri, mapping.ontology_label)
                     self.finished_mapping_set.add(ontology_entry)
 

tests/pipelines/resources/expected/consequences_snp.tsv

@@ -411,6 +411,7 @@
 17:65558594:G:A	ENSG00000168646	AXIN2	synonymous_variant
 17:6556374:C:T	ENSG00000091622	PITPNM3	intron_variant
 17:6628328:A:G	ENSG00000198920	KIAA0753	synonymous_variant
+17:6628328:A:G	ENSG00000282936		3_prime_UTR_variant
 17:6693178:A:AACACACACACAC	ENSG00000141485	SLC13A5	splice_polypyrimidine_tract_variant
 17:7224245:T:C	ENSG00000072778	ACADVL	splice_donor_variant
 17:73201307:A:G	ENSG00000166685	COG1	missense_variant
@@ -900,7 +901,7 @@
 4:120785105:A:G	ENSG00000138738	PRDM5	splice_polypyrimidine_tract_variant
 4:121847458:G:T	ENSG00000138686	BBS7	missense_variant
 4:121853086:C:T	ENSG00000138686	BBS7	missense_variant
-4:122979387:G:T	ENSG00000145375	SPATA5	splice_donor_variant
+4:122979387:G:T	ENSG00000145375	AFG2A	splice_donor_variant
 4:127881908:G:T	ENSG00000142731	PLK4	missense_variant
 4:127930737:G:T	ENSG00000164073	MFSD8	missense_variant
 4:128959797:C:A	ENSG00000151466	SCLT1	intron_variant

tests/pipelines/resources/expected/evidence_strings.json

@@ -788,6 +788,7 @@
 {"alleleOrigins": ["germline"], "datasourceId": "eva", "datatypeId": "genetic_association", "clinicalSignificances": ["uncertain significance"], "confidence": "criteria provided, single submitter", "studyId": "RCV001373139", "releaseDate": "2021-04-13", "targetFromSourceId": "ENSG00000163930", "variantFunctionalConsequenceId": "SO_0001583", "variantId": "3_52402628_G_A", "cohortPhenotypes": ["BAP1 tumor predisposition syndrome", "BAP1-related tumor predisposition syndrome", "Tumor predisposition syndrome", "Tumor susceptibility linked to germline BAP1 mutations"], "diseaseFromSource": "BAP1-related tumor predisposition syndrome", "diseaseFromSourceId": "C3280492", "diseaseFromSourceMappedId": "MONDO_0013692", "variantHgvsId": "NC_000003.12:g.52402628G>A"}
 {"alleleOrigins": ["germline"], "datasourceId": "eva", "datatypeId": "genetic_association", "clinicalSignificances": ["uncertain significance"], "confidence": "criteria provided, single submitter", "studyId": "RCV001373139", "releaseDate": "2021-04-13", "targetFromSourceId": "ENSG00000163930", "variantFunctionalConsequenceId": "SO_0001583", "variantId": "3_52402628_G_A", "cohortPhenotypes": ["BAP1 tumor predisposition syndrome", "BAP1-related tumor predisposition syndrome", "Tumor predisposition syndrome", "Tumor susceptibility linked to germline BAP1 mutations"], "diseaseFromSource": "BAP1-related tumor predisposition syndrome", "diseaseFromSourceId": "C3280492", "diseaseFromSourceMappedId": "Orphanet_289539", "variantHgvsId": "NC_000003.12:g.52402628G>A"}
 {"alleleOrigins": ["germline"], "datasourceId": "eva", "datatypeId": "genetic_association", "clinicalSignificances": ["benign"], "confidence": "criteria provided, single submitter", "studyId": "RCV001730858", "releaseDate": "2022-02-20", "targetFromSourceId": "ENSG00000198920", "variantFunctionalConsequenceId": "SO_0001819", "variantId": "17_6628328_A_G", "cohortPhenotypes": ["Joubert syndrome 38"], "diseaseFromSource": "Joubert syndrome 38", "diseaseFromSourceId": "C5561958", "variantHgvsId": "NC_000017.11:g.6628328A>G"}
+{"alleleOrigins": ["germline"], "datasourceId": "eva", "datatypeId": "genetic_association", "clinicalSignificances": ["benign"], "confidence": "criteria provided, single submitter", "studyId": "RCV001730858", "releaseDate": "2022-02-20", "targetFromSourceId": "ENSG00000282936", "variantFunctionalConsequenceId": "SO_0001624", "variantId": "17_6628328_A_G", "cohortPhenotypes": ["Joubert syndrome 38"], "diseaseFromSource": "Joubert syndrome 38", "diseaseFromSourceId": "C5561958", "variantHgvsId": "NC_000017.11:g.6628328A>G"}
 {"alleleOrigins": ["germline"], "datasourceId": "eva", "datatypeId": "genetic_association", "clinicalSignificances": ["likely benign"], "confidence": "criteria provided, single submitter", "studyId": "RCV002057147", "releaseDate": "2022-06-09", "targetFromSourceId": "ENSG00000115904", "variantFunctionalConsequenceId": "SO_0001627", "variantId": "2_39120324_C_A", "variantRsId": "rs368569135", "cohortPhenotypes": ["Noonan spectrum disorder", "RASopathy", "rasopathies"], "diseaseFromSource": "RASopathy", "diseaseFromSourceId": "C5555857", "diseaseFromSourceMappedId": "EFO_1001502", "variantHgvsId": "NC_000002.12:g.39120324C>A"}
 {"alleleOrigins": ["germline"], "datasourceId": "eva", "datatypeId": "genetic_association", "clinicalSignificances": ["pathogenic"], "confidence": "criteria provided, single submitter", "studyId": "RCV002247286", "releaseDate": "2022-06-09", "targetFromSourceId": "ENSG00000198712", "variantFunctionalConsequenceId": "SO_0001631", "variantId": "MT_7512_T_C", "variantRsId": "rs199474817", "cohortPhenotypes": ["COX deficiency", "Complex 4 mitochondrial respiratory chain deficiency", "Complex IV deficiency", "Cytochrome-c oxidase deficiency", "Cytochrome-c oxidase deficiency disease", "Deficiency of mitochondrial respiratory chain complex4", "MITOCHONDRIAL COMPLEX IV DEFICIENCY, NUCLEAR TYPE 1", "Mitochondrial complex IV deficiency"], "diseaseFromSource": "Cytochrome-c oxidase deficiency disease", "diseaseFromSourceId": "C5435656", "diseaseFromSourceMappedId": "MONDO_0009068", "variantHgvsId": "NC_012920.1:m.7512T>C"}
 {"alleleOrigins": ["germline"], "datasourceId": "eva", "datatypeId": "genetic_association", "clinicalSignificances": ["pathogenic"], "confidence": "criteria provided, single submitter", "studyId": "RCV002247286", "releaseDate": "2022-06-09", "targetFromSourceId": "ENSG00000198786", "variantFunctionalConsequenceId": "SO_0001631", "variantId": "MT_7512_T_C", "variantRsId": "rs199474817", "cohortPhenotypes": ["COX deficiency", "Complex 4 mitochondrial respiratory chain deficiency", "Complex IV deficiency", "Cytochrome-c oxidase deficiency", "Cytochrome-c oxidase deficiency disease", "Deficiency of mitochondrial respiratory chain complex4", "MITOCHONDRIAL COMPLEX IV DEFICIENCY, NUCLEAR TYPE 1", "Mitochondrial complex IV deficiency"], "diseaseFromSource": "Cytochrome-c oxidase deficiency disease", "diseaseFromSourceId": "C5435656", "diseaseFromSourceMappedId": "MONDO_0009068", "variantHgvsId": "NC_012920.1:m.7512T>C"}

tests/trait_mapping/test_main.py

@@ -6,7 +6,8 @@
 
 import pytest
 
-from cmat.trait_mapping.main import parse_traits, process_traits
+from cmat.trait_mapping.main import parse_traits, process_traits, process_trait
+from cmat.trait_mapping.trait import Trait
 
 
 def get_test_resource(resource_name):
@@ -62,3 +63,27 @@ def test_main():
     mapped_terms = {x[0] for x in output_mappings}
     curation_terms = {x[0] for x in output_curation}
     assert len(mapped_terms) + len(curation_terms) == len(all_terms)
+
+
+def test_process_trait_exact_match():
+    # Exact match with MONDO:0009061 (in EFO and Mondo)
+    trait_name = 'Cystic Fibrosis'
+    # Use our default Zooma filters
+    zooma_filters = {'ontologies': 'efo,ordo,hp,mondo',
+                     'required': 'cttv,eva-clinvar,clinvar-xrefs,gwas',
+                     'preferred': 'eva-clinvar,cttv,gwas,clinvar-xrefs'}
+    zooma_host = 'https://www.ebi.ac.uk'
+    # Don't use OxO
+    oxo_targets = []
+    oxo_distance = 0
+
+    # This should be marked as finished, as it's an exact string match with a term contained in the target ontology
+    efo_trait = process_trait(Trait(trait_name, None, None), zooma_filters, zooma_host, oxo_targets, oxo_distance,
+                              target_ontology='efo')
+    assert efo_trait.is_finished
+
+    # This should not be marked as finished, even though Zooma finds an exact match in one of its ontologies, it's not
+    # the requested target ontology and thus still needs to be curated
+    hpo_trait = process_trait(Trait(trait_name, None, None), zooma_filters, zooma_host, oxo_targets, oxo_distance,
+                              target_ontology='hp')
+    assert not hpo_trait.is_finished